MONETARY POLICY, WAGE RATES, AND PRODUCTIVITY
IN THE UNITED STATES AND UNITED KINGDOM,
1970 TO 1990
BY
PENG CHENG WEN
A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENT FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY
UNIVERSITY OF FLORIDA
1991
TABLE OF CONTENTS
page
LIST OF TABLES i v
LIST OF FIGURES v
ABSTRACT viii
CHAPTERS
1 INTRODUCTION 1
1 . 1 Introduction 1
1.1.1 Causation, Accommodation and Discipline 5
1.1.2 The Four Hypotheses 6
1.2 Episodic Change and Exogeneity n
1.2.1 Monetary Policy 14
1.2.2 The Policy Approaches, 1970s and 1980s. 17
1.2.3 Some Qualifications 18
1 . 3 The Selected Time Series 22
1.3.1 Money Demand 22
1.3.2 The Time Frames 24
1.3.3 The Data 26
2 THE CONFLICT OVER USES OF STATISTICAL METHODS.. 27
2 . 1 Introduction 27
2.2 Filtering and Episodic Change: F/S vs H/E. . 28
2.2.1 Phase Averaging 31
2.2.2 The Filtering of Trend Lines 3 3
2.2.3 H/E's EqualLength Subsample 3 4.
2.2.4 H/E and Velocity as a Random Walk 3 7
2.3 The Multiple Regression Problems 38
3 THE MONETARY INDICATOR, INFLATION RATES, AND
MONEY DEMAND. . 46
3 . 1 Introduction 4 6
3.2 The Monetary Indicator, Money Demand:
an Episodic View 48
3.3 Prices, Nominal Wage Rates and Monetary
Discipline 54
3 . 4 Statistical Results 57
3.4.1 The Trend Analysis 57
3.4.2 The Mean Analysis 66
3 . 5 Summary 71
4 MONETARY POLICY, WAGE, AND PRODUCTIVITY 73
4 . 1 Introduction 73
4.1.1 The Post Keynesian Theory of Inflation. 74
4.1.2 The RousseasWeintraub Theorem and
Monetary Policy 75
4.2 Frazer/Friedman Wage Bargaining Theory 78
4.2.1 Frazer's Analysis 78
4.2.2 Friedman ' s Analysis 82
4.2.3 The Alternative — A Restatement 84
4.3 Testing the Hypothesis 85
4.3.1 The Trend 86
4.3.2 The Mean 99
4.4 Summary 101
5 THE SAMPLE PERIODS, THE ANALYSIS OF DATA, AND
THE HYPOTHESES 106
5 . 1 Introduction 106
5.2 A Use of Conventional Method 109
5.3 A Comparison of Results 113
5 . 4 Summary 117
5.4.1 Hypothesis 1 119
5.4.2 Hypothesis 2 119
5.4.3 Hypothesis 3 121
5.4.4 Hypothesis 4 122
APPENDICES
A EXOGENOUS AND ENDOGENOUS VARIABLES 125
B METHOD OF PHASE AVERAGING 129
C THE TESTING FOR THE BEST FIT TREND LINE 131
D WAGE COMPENSATION PRICE NEUTRAL? 145
E TESTING FOR DIFFERENCE IN AVERAGE VALUES FOR
THE 1970S AND 1980S RESPECTIVELY: M" AND W~ 148
REFERENCES 155
BIOGRAPHICAL SKETCH 165
LIST OF TABLES
page
31 Maximum, Minimum, Mean and Standard Deviation
for the Indicator of Monetary Policy 67
32 Maximum, Minimum, Mean and Standard Deviation
for the Inflation Rates 68
41 Summary for Fitting Model between M" and W~
United States 95
42 Summary for Fitting Model between M~ and W~
United Kingdom 95
43 The Upper and Lower Bound Estimation between
M~ and W~ , United States 97
44 The Upper and Lower Bound Estimation between
M~ and W~ , United Kingdom 97
45 Maximum, Minimum, Mean and Standard Deviation
for the Difference between the Growth in wages
and the Growth in Productivity 101
51 Conventional Method for the relationship between
M~ and W~ , United States Ill
52 Conventional Method for the relationship between
M~ and W~ , United Kingdom Ill
Bl Income Phase Average, Income Rate of Change from
The Initial Point to the Terminal Point and Income
Growth Rate from Phase to Phase 131
Cl Summary for Fitting Model for the Trend Line W~
United States 135
C2 Summary for Fitting Model for the Trend Line W~
United Kingdom. 135
C3 Summary for Fitting Model for the Trend Line M~
United States 136
C4 Summary for Fitting Model for the Trend Line M~
United Kingdom 13 6
iv
LIST OF FIGURES
page
1la Trend Line of the Indicator of Monetary Policy,
United States
Source: Federal Reserve Bank of St. Louis 19
1lb Trend Line of the Indicator of Monetary Policy,
United Kingdom
Source: Bank of England 2
2la H/E's "EgualLength Subsample" between Money
Stock and Price for the United States
Source: Hendry and Ericsson 1990, 10 35
2lb H/E's "EqualLength Subsample" between Money
Stock and Price for the United Kingdom
Source: Hendry and Ericsson 1990, 10 36
3la Money Demand in the U.S. during the 1970s and
1980s Respectively
Source: Federal Reserve Bank of St. Louis 52
3lb Money Demand in the U.K. during the 1970s and
1980s Respectively
Source: Bank of England 53
32a Trend Line for the Inflation Rate, United States
Source: Federal Reserve Bank of St. Louis 59
3 2b Trend Line for the Inflation Rate, United Kingdom
Source: Bank of England 60
3 3 a Trend Line for the Nominal Wage Rate,
United States
Source: Federal Reserve Bank of St. Louis 62
33b Trend Line for the Nominal Wage Rate,
United Kingdom
Source: Bank of England 63
34a The Mean for the Indicator of Monetary Policy
1970s vs 1980s, United States
Source: Federal Reserve Bank of St. Louis 69
34b The Mean for the Indicator of Monetary Policy
1970s vs 1980s, United Kingdom
Source: Bank of England 70
v
41 The PriceOutputWages Connection Aggregate Demand
and Aggregate Supply
Source: Frazer 1991a, 354 80
42a Trend Line for the Difference Between the Growth
in Wages and the Growth in Productivity, U.S.
Source: Federal Reserve Bank of St. Louis 89
42b Trend Line for the Difference Between the Growth
in Wages and the Growth in Productivity, U.K.
Source: Bank of England 9
43a The Difference Between the Growth in Wages and
the Growth in Productivity, U.S.
Source: Federal Reserve Bank of St. Louis 92
43b The Difference Between the Growth in Wages and
the Growth in Productivity, U.K.
Source: Bank of England 93
44a The Mean for the Difference Between the Growth in
Wages and the Growth in Productivity 1970s vs
1980s, U.S.
Source: Federal Reserve Bank of St. Louis 102
44a The Mean for the Difference Between the Growth in
Wages and the Growth in Productivity 1970s vs
1980s, U.K.
Source: Bank of England 103
51 Comparison for Different Results between M~ and W~
in the United States, Friedman vs H/E
Source: Federal Reserve Bank of St. Louis 115
52 Comparison for Different Results between M~ and W~
in the United Kingdom, Friedman vs H/E
Source: Bank of England 116
Cl The Trend Line for W~ , as W~= a + bt + Cz
United States
Source: Federal Reserve Bank of St. Louis 137
C2 The Trend Line for W~ , as W~= a + bt
United States
Source: Federal Reserve Bank of St. Louis 138
C3 The Trend Line for W~ , as W~= a + bt + Cz
United Kingdom
Source: Bank of England 13 9
C4 The Trend Line for W~ , as W~ = a + bt
United Kingdom
Source: Bank of England 140
vi
C5 The Trend Line for M" , as M"= a + bt + Cz
United States
Source: Federal Reserve Bank of St. Louis 141
C6 The Trend Line for M~ , as M~= a + bt
United States
Source: Federal Reserve Bank of St. Louis 14 2
C7 The Trend Line for M~ , as M~= a + bt + Cz
United Kingdom
Source: Bank of England 143
C8 The Trend Line for M~ , as M~= a + bt
United Kingdom
Source: Bank of England 144
VII
Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy
MONETARY POLICY, WAGE RATES, AND PRODUCTIVITY
IN THE UNITED STATES AND UNITED KINGDOM,
1970 TO 1990
By
PENG CHENG WEN
May 1991
Chairman: Dr. William Frazer
Major Department: Economics
This dissertation appears against a broad background of
literature in economics concerning uses of statistical
methods, and Milton Friedman and Anna Schwartz's Monetary
Trends in the U nited States and the United Kingdom (1982) .
At the time of the publication of Monetary Trends , Charles
Goodhart took note of the unusual use of statistical methods
in a review article, and afterward econometricians Hendry and
Ericsson attacked the Friedman and Schwartz approach.
William Frazer also had pointed to Friedman's departure
from the econometrician' s uses of statistical methods in a
1983 article and undertook further study of Friedman's uses of
statistical methods. This became a part of Frazer 's
discussion of Friedman's use of statistical methods in Power
and Ideas (1988) . He saw that Friedman not only offered an
alternative way of analyzing the statistical data but that he
viii
had a total alternative analytical system which extended to
the use statistical methods.
Against this background, we undertook analyses of data
for both the U.S. and the U.K., asked specific questions, and
offered specific hypotheses as we considered the data for the
1970s and 1980s decades. At the juncture of the two decades,
the policies of the U.S. and U.K. take redirection and appear
as what Frazer calls "The Big UTurn." In broad outline the
policy moves from monetary accommodation to monetary
discipline.
Drawing on Friedman's uses of analysis and statistical
methods, we see these two distinct decades of policy as
episodes which impact on the time series we undertook for
study. Along this route, some further support arises for a
prior hypothesis namely:
Hypothesis 1: The United Kingdom and the United States
have in common the same determinants of the money demand
functions.
The additional hypotheses we presently address are as
follows:
Hypothe sis 2 : Nominal wage rates adjust more readily in
the presence of monetary discipline.
Hypothesis 3 : Wage rates are determined by productivity
and market structures irrespective of monetary policy.
Hypothesis 4: Standard econometric methods are
appropriate for analysis of the time series we deal with,
the hypotheses we confront, and the treatment of episodes
of the kind we encounter for the decades of the 1970s and
1980s.
Given the approach we adopt, the statistical results reported
support hypothesis 2, but not hypotheses 3 and 4.
ix
CHAPTER 1
INTRODUCTION
1. 1 Introduction
William Frazer has identified alternative analytical
systems which appear in economics over the postWorld War II
years (Frazer 1988, 453456; 1991a), and which share some
claim to being "positive economics" as Friedman discussed it
(Frazer and Boland 1983; Friedman 1953). In addition, He has
traced out alternative ways of treating probability as it
enters into the alternative systems (Frazer 1988, 6768) .
Looking only at the analytical systems, Frazer narrows the
treatment to Friedman's system and to the Keynesian/post
Keynesian system. He focuses the analysis on policyoriented
guestions and approaches associated with Keynes's economics
and Friedman's economics. 1
Injecting the use of probability, Frazer finds that
Friedman adopts a Bayesian (or subjective) probability in his
analytical approach, and that the Keynesians come closest to
Frazer 's most recent work in this respect is A European
Monetary System (1991b) . Drawing on the material in this
1991b study and a wide range of previous policyoriented
studies (e.g., Frazer 1967; 1973; 1978; 1988), Frazer
concludes in the 1991b study that the Keynesian/post
Keynesian system and Friedman's are the only alternatives
that have any meaningful visibility at policy levels in
the 1980s and in the 1990s as plans for a new European
monetary system appear.
2
identifying with what Milton Friedman and Anna Schwartz
(hereafter F/S) called "the prevailing fashion in econometric
work" (Frazer 1988, 79; F/S 1982, 629). In "the prevailing
fashion in econometric work" the more classical probability
enters. Statistical results are obtained as if the data under
analysis were drawn from an unchanging universe where episodes
play a very secondary role. Often the only role the episodes
play appears as a dummy variable in assessing the information
obtained by the use of the statistical methods.
In Friedman's case, episodes play a main role in moving
the time series about, and the behavioral units ( the firms,
households, and individuals) are in the position of learning
and adjusting to them as behavioral units move through time.
Further, Friedman's search with respect to the time series was
for a stable relation, so he consequently set about
distinguishing between information in the time series that was
transitory and information that was more permanent. The
major, first study along this line was A Theory of the
Consumption Function (1957), from which comes "the permanent
income hypothesis." "Permanent" became synonymous with
"trend" (the "trend" component in the data) , as F/S undertook
such works as the Monetary History (1963) and Monetary Trend
(1982) .
In even these early studies, adjustments to the time
series for "episodic change" began to appear in the time
series spanning a hundred years or more. As studies continued
3
and matters evolved, the "episodic change" dimension became
more complicated. Particularly it did so as the U.S. and the
U.K. undertook particular policy control actions which may be
said to be based on Keynesian economics. The major sorts of
episodic changes Friedman started to point to as the post
World War II years drew on were with respect to "psychological
time," 2 and changes in the structure underlying the formation
of inflationary expectations which Friedman dated from the mid
1960s in both the U.S. and the U.K. (F/S 1982, 569573).
In general terms episodes became equated with "exogenous"
(to use a technical term) influences, and debate arose over
the treatment of monetary policy as being an outside (or
exogenous) event. It did so as a part of what became the
causation debate which took up uses of statistical methods
(Frazer 1973, 126129) and which Lord Kaldor extended to
institutional considerations (Frazer 1983; 1988, 9798, 545,
729730; 1991a, chap. 3).
Frazer, too, picked up on Bayesian probability, learning
on the part of the behavioral entities, and monetary policy.
Initial discussion appeared in 1973 (sect. 8.2) and a more
formal piece appeared in 1978 (Frazer 1978) .
"Psychological time" (Frazer 1988,742) refers to an
episode (or a policyinduced episode) such as the stock
market crash of 1987 (Frazer 1988, 683686) where the
behavioral units form expectations by examining the
current condition and analogizing it to a past similar
episode such as the stock market crash of 1929.
4
Starting with a 1984 interview with Friedman, Frazer
sought to give more order to Friedman's uses of Statistical
methods (Frazer 1988, chap. 18; and Frazer and Sawyer 1984). He
later identified Friedman's early monograph titled "The
Interpolation of Time Series by Related Series" (1962) as
representative of this orientation.
Along this line of inquiry what we end up with are the
following points: (1) Friedman's search was for stable
phenomena and not "flat out" stability in the relation (Frazer
1984) . (2) The impacts of nonrepetitive changes on the time
series must be separated from those that can be said to be
repetitive. (3) Episodes impact on the time series in all of
the main time frames we adopt, namely, the very short run, the
transitory period (or short cycle emphasized by Friedman) , and
the long run (i.e., Friedman's permanent or "trend"
components) . They do so particularly in the highly volatile
states of the 1970s and 1980s. (4) Central bankers in
particular are interested in reacting to liquidity shifts
bearing on the very short run (e.g., the stock market crash of
1987), the stabilization of business conditions (i.e.,
transitory change) , and the attainment of economic growth
(i.e., Friedman's permanent components). (5) The decades of
the 1970s and the 1980s are decades in which monetary policies
of a very different sorts were imposed on the time series from
outside of the system of time series, particularly those for
prices, wages, and productivity. With respect to the latter,
5
we take up the definitions below of "monetary accommodation"
and "monetary discipline."
A further aspect of Frazer's treatment of alternatives
analytical systems, which will be of later interest, is that
Friedman's system has its own microeconomic foundations.
Along this line, Friedman rejects the theories of market
structures which appeared since the 1930s (Frazer 1988, chap.
9) . He goes back to Keynes of the 1920s and to some extent
Keynes of the 19 3 0s, but he proceeds very differently with
special time frames and all, and very differently from the
Keynesian/PostKeynesian alternative. As we have stated, the
special time frames include the long run (the trend component
of the time series) , the transitory (or cyclical) component of
the series, plus there is a place for episodic changes. To be
sure, monetary policy itself may be viewed as an outside event
(as "exogenous" to use a technical term) which receives
attention in Appendix A.
1.1.1 Causation, Accommodation and Discipline
The issue of causation in this monetary, statistical,
policyoriented economics, spans over decades and reduces to
the guestion of whether money growth (M°) causes income growth
(Y°) with feedback or whether the income growth causes the
money growth ( AY — > AM . ) (Frazer 1973, 125131; 1991a,
sect. 3. 5; and 1991b),. It reduces to a very specific set of
policy operations where we have the central bank controlling
the money stock independent of fiscal policy and interest
6
rates, a la Friedman, or whether fiscal policy (defined as a
deficit) becomes the principal control means with the central
bank simply accommodating the control.
The distinctions between Friedman's alternative and the
Keynesian/PostKeynesian system goes beyond the time frames,
the uses of statistical methods; it extends to issues of
causation, and the role of the government and central bank
operations as most recently addressed by Frazer (1991b) .
These differences are what we portray as monetary
accommodation, (AM * 100)(Y/M)>0, and monetary discipline,
(AM * 100) (Y/M)<0. In this distinction AM signifies monetary
acceleration, AM signifies deceleration, and Y/M is income
velocity of money. It indicates the extent of the final
impact on spending of the acceleration (or deceleration) .
There is more, but this introductory review serves only
to point out that the present work focuses more narrowly on
Friedman's approach, the Keynesian /PostKeynesian
alternative, and particularly undertakes to engage in
statistical analyses of data which we identify with Friedman
as opposed to "the prevailing fashion in econometric work"
(Frazer 1988, 79; F/S 1982, 629). Along this line, we may
point to hypotheses 1 and 4 below.
1.1.2 The Four Hypotheses
The present study are four hypotheses, they are:
Hypothesis 1: The United Kingdom and the United
States have in common the same determinants of the
money demand functions (F/S 1982, sect. 5.4).
Hypothesis 2: Prices and nominal wage rates adjust
more readily in the presence of monetary
discipline. (We recall that Keynes pointed to
"sticky wage" in the 1920s and based his General
Theory on the notion of a wage standard i.e., that
wages would remain tied to productivity growth as
total spending was managed to achieve Keynesian
full employment.) 3
Hypothesis 3 : Wage rates are determined by
productivity and market structures irrespective of
monetary policy. [This we associate later (sect.
4.1.2) with a theorem due to Sidney Weintraub and
Stephen Rousseas. We use it in reference to the
post Keynesian theory of an "endogenous" money
supply. ]
Hypothesis 4: Standard econometric methods are
appropriate for analysis of the time series we deal
with, the hypotheses we confront, and the treatment
of episodes of the kind we encounter for the
decades of the 1970s and 1980s. [This statement
gains further significance from the reliance of
these methods on sampling from an unchanging
universe over a long period of time. The
analytical problems here are associated with what
Frazer calls " the separation of effects " problem
(1991a sects 2.4b and 2.5) and with the
classification of variables as endogenous.]
The first of these hypothesis was very much supported by
F/S's Monetary Trends (1982) . A Major result of the findings
giving rise to the statement was that no conditions special to
one of the countries needs to be brought into discussion as
far as the empirical findings are concerned.
Hypothesis 2 gains its importance for having been
introduced by J. M. Keynes in the 192 0s, in the form of the
issue of sticky prices and the special price called the wage.
As built into Keynes's General Theory (1936), the idea is in
We see this as being widely accepted, but Sir John Hicks
offers discussion of this position (Hicks 1983) .
8
two parts; that nominal wages (and thus prices) do not adjust
downward in the presence of inadequate demand for production
(say, because of a shift in liquidity) ; and that production
adjust downward rather than price. Were the prices to adjust,
demand for the current output would be restored and full
employment production attained. With unemployment in the
presence of what was perceived as a failure of monetary
policy, the economicpolicy solution to unemployment was in
terms of fiscal policy. The failed view of monetary policy
pertained to the Keynes/Keynesian monetary policy linkages
where reliance focused upon interest rates as the control
variables (Frazer 1991a chaps. 3 and 4) . The combination of
the perceived failure of monetary policy, and the fiscal
policy on the positive side is where we get monetary policy
with the purpose of accommodating price increase and deficit
spending, such as we encounter in the U.S. and the U.K. in the
1970s, and at the hands of Lord Kaldor (1982, 4260).
As reviewed by Frazer, Friedman's orientation toward
monetary policy was taken up by Ronald Reagan in the United
States as a part of what was called "supplyside economics"
(Frazer 1988, chap. 16). The same control arrangement was
called "monetarism" in the U.K., where Friedman's influence
was also felt (Frazer 1988, chap. 14 and 15). The idea in both
cases was to use monetary policy to tame inflation (and hence
in general bring about a downward adjustment in the inflation
rate which had gotten built into pricing policy and wage
9
contracts in the 1970s) . Thus, the economics and the
political positions found in the U.S. and the U.K. in the
1980s are linked togather.
Hypothesis 3 gains its place in the present study because
of its strong link with the monetary accommodation view which
we see as principally Keynesian/postKeynesian. In contrast
to Friedman's monetarist view, the explanation for inflation
resides in theories of market power, and no attention at all
is given to monetary matters as a means of stabilizing the
price average. So once again the distinction we encounter
between the decades of the 1970s and 1980s, and monetary
accommodation and discipline extend readily to the attention
we give hypothesis 3 .
Hypothesis 4 gains its place in this study because
Friedman adopts a Bayesian (or subjective) probability in his
analytical approach, and the Keynesians come closest to
identifying with "the prevailing fashion in econometric work"
method. In Friedman's case, episodes play a main role in
moving the time series about (sect. 1.1). On the other hand,
the "the prevailing fashion in econometric work" approach
obtains statistical results as if the data under analysis were
obtained from an unchanging universe and episodes play a very
secondary role.
In confronting "the prevailing fashion in econometric
work," special analytical problems in analyses of data arise.
One is that dealing with solutions for variables internal to
10
the model in term of variables called "exogenous" which are
outside of the model, and another is that with the separation
of effects in the multiple regression case. As will ne taken
up in the next chapter (sect. 2.3), the separation of effects
comes down to whether the socalled independent variables in
a multiple regression are in fact independent, plus there is
the Learner problem of setting limits on the true regression
coefficients in such models (Learner 1985) .
With these analytical matters in mind, Friedman offers a
different way to proceed. His uses of method are simple and
indirect and offer several main prospects. One of these is
the prospect of allowance for the impact of episodes on the
time series. Closely related is the prospect that information
is being obtained from a changing universe. Also there is the
prospect that no variable is entirely "endogenous" or
"exogenous" by the definitional standards the econometricians
have set (appendix A) .
We will not deal with all of this in great detail, since
it appears elsewhere (Frazer 1988; 1991a and 1991b), but we
bring to the forefront the presence of "episodic change" in
the time series. On the one hand, we see it as a problem in
the use of "the prevailing fashion in econometric work"
(Frazer 1988, 79; F/S 1982, 629), and, on the other, the use
of methods found in Friedman's approach elevates the
importance of episodic changes and highlights sampling from
different universes.
11
1.2 Episodic Change and Exogeneity
Substantial changes in time series may occur from
developments outside the system of equations which econometric
technique made fashionable. This encountervailing view
appears along two lines: (1) the main regime shifts we point
to and (2) the approach Friedman took to monetary policy as an
"exogenous" variable (appendix A) . Indeed, he even made the
analogy to "helicopter" money in explaining his view of money
whereby routine drops of twenty dollar bill on the community
by the helicopter (Frazer 1991a, chap 3; 1991b, sect. 2.1c)
would lead to additional spending (hence, AM >AY°, or
conversely AM > AY ) .
In the first instance, there are "shocks" impacting on
numerous time series of the type we take up, although we do
recognize that some "shocks" may not be entirely independent
of the time series themselves. The sorts of "shock" Frazer
cites most often may be illustrated by the following list: (1)
the political regime shifts such as occurred in the United
States and the United Kingdom with Reagan and Thatcher,
respectively; (2) the numerous announcement in the 1970s,
first by Richard Nixon and then Jimmy Carter, about price
controls; (3) President Nixon's announcement in 1971, about
the U.S.'s intention of no longer supporting the U.S. dollar
with gold; (4) oilcartel pricing which first appeared in
1973; (5) the Iranian crisis and related oil pricing in 1979;
(6) Ronald Reagan's confrontation with the air traffic
12
controllers in the spring of 1981; (7) Margaret Thatcher's
confrontation with the British coal miners on successive
occasions; (8) Reagan's attacks on the Federal Reserve in the
early 1980s to bring about some of the results we see as
monetary discipline; (9) news reported through the U.S.
Treasury's secretary in 1985 and on later occasion to the
effect that the U.S. would allow the dollar to decline in
price in the foreign exchange market rather than to pursue
further deceleration of inflation rates; (10) the stock market
crash of 1987; (11) Reagan's and Greenspan's announcement at
the October 1987 date that Federal Reserve would not repeat
the mistakes of the past and the subsequent appearance of open
market purchases in New York by the Fed; and (12) the
privatization of British stateowned industry in the second
half of the 1980s; (13) the Iraqi invasion of Kuwait in 1990.
In our examination of episodes, we adopt Friedman's view
that monetary policy itself is an impact on the economic
system from outside (i.e., the view of AM — > AY ), and
utilize the present monetary discipline measure [i.e., (
AM°*100) (Y/M) ] . Nevertheless, we do not expect that the
effects of the monetary discipline are entirely independent of
the resolve Reagan and Thatcher provided. Reagan's firing of
the air traffic controllers in 1981, and Thatcher's
confrontations with the coal miners, must have helped by
reinforcing the monetary policy in bringing about the price
and wage changes we point to.
13
We ask, following the introduction of the presence of
episodic change, "What is the information contained in a time
series?" In answer, we expect some of it to be purely non
repetitive, as where NOW accounts were included in the money
stock measure $M1 in the early 1980s (Frazer 1988,665687);
some of it may represent changes in business conditions (where
income, Y, varies about a more permanent income measure, Y p ) ;
and some of time series component may be permanent (e.g., Y p ) .
We have chosen to analyze data for the decades of 1970s and
1980s, where we see the main difference between the decades as
being of an episodic nature. To be sure, we see these two
decades, which we presently juxtapose, as being sufficiently
differentiated in terms of outside forces that they provide
the prospect for significant differences in the time series
drawn from the respective decades.
So there are two major features of the work at hand: (1)
the testing of the hypotheses themselves (sect. 1.1.2) and (2)
study bearing on the content of information contained in the
time series data. Indeed, there is in the latter case the
twofold matter (a) of a given series (say, Y) containing
several classes of information and (b) that of obtaining
information from essentially different universes. It is this
emphasis on outside forces as dominant ones in both instances,
(a) and (b) , that is at odds with the fashion in econometric
practice which F/S pointed to. The presence of the outside
forces in the 1970s and 1980s leads to a rejection of the
14
"prevailing fashion in econometric work" for use in dealing
with the class of economicpolicy problem with which we are
concern. As pointed out, an underlying theme of the 1980s
goes back to J.M. Keynes of the 192 0s, notably that wages are
sticky and fail to adjust. The 1980s counterpart to this is
that wages adjust differently in an era of monetary
accommodation visavia one of monetary discipline (Frazer
1988, 649, 668670). In the 1970s we have what Frazer calls
the Keynesian era (monetary accommodation and all as imposed
via government) and in the 1980s we have monetary discipline
to a reasonably significant extent (Frazer 1988, 649670).
1.2.1 Monetary Policy
As we have already indicated, monetary policy in the
1970s is very Keynesianoriented in that monetary policy is
accommodative of fiscal policy (defined as a deficit) ,
inflation, and wage increases. To the extent that inflation
arises in the Keynesian/PostKeynesian analytical system it is
viewed as a marketstructures problem — oligopoly,
administrated prices. Further, power theories of inflation
generally enter (Frazer 1988, 208209, 229, 545546).
Parallel to this marketstructures view of inflation, the
control of inflation is sought, through price controls where
efforts are made to select for the purpose of control firms
which "administer prices."
In the 1980s, the marketstructures/pricecontrol
approach was dropped, monetary policy was viewed as the
15
principal means of controlling inflation. Money and credit
aggregates gained ascendancy as a focal point for controlling
and taming inflation. Even as this approach came to be adopt,
however, there were special analytical and operational
problems at the respective central banks involved in the
policy execution. These center primarily about the
following: (1) difficulties in targeting a principal monetary
aggregate, $M1, when new classes of deposit liabilities at the
commercial banks get included in the principal aggregate $M1
(Frazer 1988,655657); (2) difficulties encountered with the
actual execution of policy because of inadequate accounting
control arrangements, or because of past practices and
traditions at the operations level 4 ; and (3) the dramatic
change, in the U.K. case, where the U.K. undertook the
privatization of previously government owned enterprises in
the second half of the 1980s. As the 1980s closed (Frazer
4
In his 1991b monograph (sect. 1.2), Frazer offers a
principal hypothesis, namely:
Traditions, operating procedures, and
accounting controls influence the choice
of economic theory on which government
bases its central banking and financial
markets policies. (Notes: Sharing the
determinants of the money demand function
and the workings and economic laws as
reported in F/S's Monetary Trend are one
thing. The choice of a theory on which
to implement policy is another. It may
have not only political overtones but
dependence on accounting arrangements,
for accounting control purposes, and on
past practices and traditions. ) (1991b,
6).
16
1991b, chap. 4), the public in the U.K. had acquired enormous
amounts of liquid assets in the form of marketable shares in
the enterprises, and this group ran up considerable debt as
the public appeared to dissave in current income terms and the
government became a net saver in such terms (i.e., the
government ran a surplus in its budget and retired outstanding
debt) .
All of this complicates the analysis of the demand for
money and the execution of monetary policy, the attainment of
targets for reserves, the money stock and bank credit in a
refined sense. Even so, for the present purpose these
complications may be filtered out of the data by focusing on
Milton Friedman's uses of statistical methods, and the general
trend of monetary policy in the respective decades.
As indicated, the overall role for monetary policy was
different for the respective decades in both the U.S. and the
U.K.. There is what Frazer has called the "iregime" and the
"Mregime" (Frazer 1991b, sect. 2.3, 2.4 and 2.5). 5 And
The respective regimes have many dimensions, and
have been extensively written about (Frazer 1988) . In
brief, the "iregime" label encompasses a traditional
banking and Keynes ian view which harks back to the days
when central banks' sole means of intervening in the
money and credits markets was through the discount rate.
"The interest rate" is also at the center of policy in
the analytical constructs passed along by the Keynesians.
This remains true even when we recognize that Britain's
J. M. Keynes of 1930s prominence showed special interest
in the discovery of open market operations at the Federal
reserve Bank of New York.
The Mregime contains interpretations and analyses
which substitute for the "iregime" interpretations and
17
although a full Mregime orientation may not have been
feasible in the U.K. for reasons given by Frazer, we may
nevertheless view monetary policy as actually pursued in terms
of the principal monetary aggregates (Ml in the U.S., and M3
in the U.K.) . This is essentially what Milton Friedman did in
the Monetary History (1963) and Monetary Trends (1982)
volumes with Anna Schwartz. In other words, irrespective of
whether the central banks attempted to report monetary (or
money and credit) policy in interest rate or moneyaggregate
terms, the measure Friedman used was the money aggregate (its
time rate of change) . This followed essentially from his
monetary theory and his total analytical system.
1.2.2 The Pol icy Approaches. 1970s and 1980s
Establishing the extreme differences in the policy
approaches to the two decades was a task Frazer undertook in
his Power and Ideas (1988, chaps. 8, 14, 15 and 16). At
approximately the turn between the two decades, a reverse set
of policies enter in both the U.S. and the U.K. under Ronald
Reagan and Margaret Thatcher, respectively. The policies for
analyses. Stated this way, we associated it mainly with
Friedman's monetarism, and no other views of monetarism.
This view is strongly rooted in a U.S. tradition of
central banking, where OMO's of a special sort play a
main role and where the means of accounting for the money
and credit aggregates are fairly straight forward (Frazer
1991b, sect. 2.4). This statement is thought to apply
despite two developments: (1) the extension of F/S's work
to include the U.K. and financial interrelationships
between the U.K. and the U.S.; and (2) despite reliance
by different countries, and particularly the U.K. and the
U.S., on monetarists ideas to tame the almost worldwide
inflation of the 1970s.
18
the most part were to contain and tame inflation by monetary
means; to free up the private sector by changing tax
structures along incentivist, minimumgovernmentinterference
lines; and to eliminate and avoid direct price controls. We
see the respective sets of policy— symbolized by the monetary
accommodation in the first instance and discipline (inflation
taming) in the second — as mainly occurring from outside of the
time series and as being imposed on them. Said differently,
we see the monetary role as a primary one shaping the time
series more generally, and conseguently, we proceed very much
with what may be called a Friedman/Frazer approach to the
analysis of the time series.
The money and credit aggregates approach is quite
compatible with F/S's approach in Monetary Trends , and the
hypotheses stated about (sect. 1.1.2).
1.2.3 Some Qualifications
Frazer has pointed to some qualifications. Most notably,
the monetary policy we indicate and define may be inadvertent
on the part of officials in some respects in the 1970s in both
the U.S. and the U.K.. To be sure, the interest rate was
viewed as the principal control variable on the part of those
at the U.K. Treasury and at the Bank of England, even as the
reporting of policy in money and credit aggregate terms was
debated and imposed on the central banking authorities. At
the end of the first decade and the beginning of the second,
19
%
20
10
2
o
o
o
s
o

10 
20
H h
_) 1 1 1 i
H 1 1 1 1 1 1 1 1 1
1968 1972 1976 1980 1984
Year
1988 1992
Figure lla Trend Line of the Indicator of Monetary Policy,
United States
20
\
o
o
H
o
10
1968
1992
Figure 1lb Trend Line of the Indicator of Monetary Policy,
United Kingdom
21
a reverse in policy came to both the U.S. and the U.K., but
there were problems of implementation, such that old ideas
about interest rates and central banking never fully
disappeared from the center of policy considerations.
These Frazer has treated as distinct iregime and M
regime approaches. He has also noted that the Bank of England
in particular was not able to fully move along Mregime lines
at the policy level for several reasons. For one, accounting
control and market intervention practice and traditions were
inadequate for the task, 6 and for another, monetary and
treasury officials in the U.K. had only a very Keynesian/i
regime view of Friedman's monetarism, even as they attempted
desperately in the first half of the 1980s to target the money
aggregate sterling M3 . Missing targets and losing public
confidence, they reverted to iregime notions and press
reporting along such lines, even as they followed fairly
disciplinary policies.
Viewed overall for the respective decades for both the
U.S. and the U.K., we illustrate the distinct differences in
figures lla and 1lb. There we see the trend of the monetary
indicator for the 1970s (actually 1970:1 to 1979:1V), the
trend for the 1980s (1980:1 to 1989:11) , and the turn from one
period to the other for the United States and United Kingdom,
respectively. We see both in the U.S. and the U. K. that the
trend line for the monetary indicator has a positive slope in
See note 4 above.
22
the 1970s. We call it monetary accommodation. Moving into
the 1980s, as shown in both figures 1la and 1lb, the trend
line for the monetary indicator takes on a negative slope. We
call it monetary discipline. We do not attempt to enter into
these matters here, except to point to them and note the
indicator of accommodation and discipline, irregardless of
policy intentions on the part of the banking officials.
1.3 The Selected Time Series.
The principal time series we select for analysis cover
the 1970s and 1980s decades. The series include those for the
money stock (Ml in the U.S., and M3 in the U.K.), price
indexes, wage rates and productivity. Attention is given to
the 1970s and 1980s principally because of the roughly egual
time periods and because each decade is characterized by a
very different approach to policy. Not only do we have the
broad Keynesian and monetarist difference, but we have the
difference between monetary policy defined in terms of the
principal monetary aggregates for the respective countries
during the 1970s and the 1980s decades (Frazer 1988, 648651,
669672) .
1.3.1 Money Demand
Following Friedman's use of methods and taking note of
his "primal eguation," 7
Drawing on work undertaken with Kim Sawyer (Frazer 1984)
Frazer defines "primal equation" as follows:
A primal equation is one which can be
estimated separately from the other equations.
23
M/P = /(Y/P, w; ...; u) (1.1)
Here M is the money stock, P is a price index, Y is income, w
is a measure for liquidity in relation to wealth, the dots
represent four expected rates of return, and u is a "catch
all" variables for secondary influences. Of present
significance, the terms M/P and Y/P yield the Cambridge k (or
the inverse of income velocity) , namely,
M = kY, or k = M/Y (1.2)
where k is the demand for money (k = 1/V, V = Y/M) . In
treating income velocity as money demand, and the money stock,
M, as a control variable, Friedman is proceeding to deal with
the "identification problem" (Frazer 1988, 543). As Frazer
pointed out, his approach to this is also different from that
found in the " fashionable econometric method." The demand
for money (i.e., velocity, V) changes for a variety of
reasons, which we avoid restating, 8 but principally we see it
in the decades at issue as a response to changes in expected
inflation (or deflation) . Moreover, the velocity at any given
time is a measure of the extent of the impact of monetary
acceleration or deceleration, AM (or AM ) on spending. It
It is predetermined; it is an equation which
comes first and from which other relationships
in the economy follow (Frazer 1988, 798 n20) .
The definition of "money" itself as stated by Keynes and
accepted by Friedman is indicative of the range of things
that may influence it. For further statement and the
definition, see Frazer (1991a, chap. 4) .
24
enters the indicators for monetary accommodation and
discipline (sect. 1.1.2).
13.2 The Time Frames
The distinctions as to time frame are (1) the very short
run (a point of market intervention with much attention to
market adjustments in response to events, and policy actions,
inactions and pronouncements), (2) the short cycle, and (3)
secular trend. Particularly in the latter instances, as we
mentioned above, we see Friedman distinguishing between
transitory and relatively permanent changes in the time
series. For income (or GNP) these main components of the time
series may be denoted YY p , and Y p , where the transitory
component is the difference between the actual data for income
and a trend component called permanent income (Y p ) , and where
the permanent component is the trend obtained by a method of
phase averaging (appendix B) in reference to phases in the
transitory component. In all instances, with respect to the
time frames, episodes may enter to give rise to what F/S
called "episodic change."
Episodes are events reported in the news, such as
illustrated above (sect. 1.2). Quite obviously, episodes are
outside of the usual economic model and indeed offer the
prospect of affecting the economic system. There is no
problem in saying that they are external to the economy's
ordinary functioning, and indeed, the issue arises most
25
vividly where Friedman treats monetary policy as exogenous and
as impacting on business conditions (say causation AM — > AY) .
Statistical and other controversy ensued over this, and
reverse causation arguments appeared. All of that is not so
much the point, however, as the fact that Friedman treated
monetary policy as exogenous, along with a host of other
events that may impact on time series of the sort he studied.
Uncertainty over the outcomes of the events, episodes, policy
action, and all entered along with matters captured by
reference to "psychological time" (note 2 above; Frazer 1988,
731; F/S 1982, 568569, 358).
So Friedman went about the search for a stable
relationship (or, more accurately, stable phenomenon) , in a
world where was impacting on the time series. The idea was to
separate out changes that were less permanent, and focus on
those that were more permanent after adjusting the time series
for the episodic part. Great effort was made on filtering
out information in the time series to focus on parts of it, in
a world where Friedman did not expect "fashionable"
econometric technigue to work. In his approach he ended, as
Frazer points out, in confronting those components in the time
series data that have traditionally interested policy makers —
shortrun crises; the stabilization of the short cycle;
smooth, less disturbed economic growth.
All of the above are present in the methods we adopt and
in the approach we take.
26
1.3.3 The Data
Our basic data are quarterly time series for (a) money
stock, (b) gross national product, (c) output per hour of all
persons, (d) hourly compensation, and annual time series for
(e) the consumer price index, and (f) the GNP deflator. The
data for the United State come from National Economic Trends
and Monetary Trends . both of which are published by the
Federal Reserve Bank of St. Louis. And the data for the
United Kingdom come from Long Run of Monetary Data which is
produced by the money & banking aggregates group, financial
statistics division, Bank of England and Economic Trend which
is produced by the Royal Central Statistical Office. And, of
course, we have supplemented the basic series in generating
new series for analyzing particular problems. Since the
present focus upon comparisons between the 1970s and the
1980s, we starts series with 1969 and terminates the series
with 1989.
CHAPTER 2
THE CONFLICT OVER USES OF STATISTICAL METHODS
2 . 1 Introduction
Although it should have been apparent that Milton
Friedman was proceeding differently from the econometricians
in his uses of statistical methods from the time of his 1957
publication titled A Theory of the Consumption Function , this
difference did not appear to be apparent to some economists
and econometricians until the publication of Friedman's and
Schwartz's Monetary Trends (1982). At that time Charles
Goodhart took note of the matter in his review article for the
Journal of Eco nomic Literature (1982, 15401551). Shortly
afterward econometricians David F. Hendry and Neil Ericsson
(Hereafter H/E) took up the sordid task of employing
"fashionable work of econometrics" to the analysis of U.K.
data in a series of papers in which they attacked the F/S
position (H/E 1984; 1989; 1990). The H/E effort went through
stages as they appeared at various conferences (Frazer 1988,
737), and by July 1989 they produced a copy titled "An
Econometric Analysis of U.K. Money Demand in Monetary Trends
in The United States and The United Kingdom by Milton Friedman
and Anna Schwartz" (H/E 1989) . Also, a further paper
27
28
analyzing time series for modeling the demand for money in the
U.K and the U.S. appeared in 1990 (H/E, 1990).
Indeed, H/E's papers provide an opportunity to further
focus on the issues over the uses of methods and to illustrate
the alternative they rely upon. The issues in a larger
context have appeared under such labels as "big models and
small models" in the past (Frazer 1973, chap. 14). However, we
see the issues as narrowing to three in number, as we proceed
with an assessment of the alternatives, particularly as
represented by H/E. The principal issues are (1) Friedman's
vision of time series, both as it relates to (a) the impact of
episodes on the data, and (b) his search for a stable (i.e.
repetitive) relation; (2) Friedman's approach to filtering
information out of the time series in order to focus on
information that interest him in the rearch for a stable
relation; and (3) his use of simple method, with attention to
the bounds on the true regression coefficient. The third is
related to the "Learner problem" (Frazer 1988, 750).
2 .2 Filtering and Episodic Change: Friedman vs H/E
Friedman's uses of statistical methods which were present
all along were not the main focus of critics' attention from
the early 1950s to the early 1980s. Instead, attention was
directed toward the importance of the money stock as a
variable. However, this was not so after the Monetary Trends
was published, as indicated by Charles Goodhart ' s article on
Monetary Trends (Goodhart 1982) . There we encounter Goodhart
29
commenting on F/S's forms of data adjustment. He objected to
the separation of trends and cycles, noted incongruity with
econometrics, and said that F/S presented evidence in an
idiosyncratic manner. Goodhart then expressed concern about
the "adjustments and manipulations imposed by F/S on their raw
data before testing. (1982, 1541)" Essentially, he pointed
to the use of "phase averaging" on the part of F/S. Yet, F/S
said that they proceed "indirectly"; that they examined
variables a few at a time with reference to hypotheses
generated by the theory; that their approach "yields insight
that cannot be obtained from the more sweeping approach (the
prevailing fashion in econometric work)" (Frazer 1988, 79; F/S
1982, chap. 2, sect. 6.2, 629).
Such divergence in the uses of methods as reflected in
Goodhart' s comments and Friedman's uses appeared in fragments
of the literature with respect to structural eguations
methods, and the Cowles Commission at the University of
Chicago (Frazer 1988, 6887). But in 1983 and '84 new charges
surfaced against F/S in the study by David Hendry and Neil
Ericsson of Oxford University's Institute for Econometrics and
Statistics and of Nuffield College. In their series of
papers, beginning in 1983 and extending to the most recent
paper dated July 1990, they picked upon what Goodhart pointed
to — "adjustments and manipulations imposed by F/S on their raw
data before testing."
30
In order to reject F/S's claims, H/E emphasized on two
issues — (1) "phase averaging" and (2) velocity as a random
walk. In doing so, they went back to the annual raw data,
because they regarded the F/S adjustments to the data and the
phase averaging as limiting the information in the data
analyzed. However, in taking this line of criticism, they
neglected giving attention to the reasoning behind F/S's use
of phase averaging. Of course, the reasons for "phase
averaging" were (1) to aid in fitting a trend and at the same
time determine beginning and terminal points for a period, (2)
to facilitate the separation of components of information
contained in a time series, and (3), as Friedman said later,
to highlight "one class of information" and "avoid its being
diluted by a class of information relevant to a question we're
not trying to answer" 1 (Frazer and Sawyer 1984) .
On the statistical problem of separating the
components which in fact concern separate time frames
for a relevant economic theory, Friedman said:
The problem is that a set of data contains
information about more than one question, and you want to
eliminate information about questions you are not
interested in. This is in order to concentrate on
information about questions you are interested in ....
Now if I had a perfect cycle, if I had a sine curve,
or alternatively, if I had a perfect theory of the cycle,
it might be possible to analyze the secular question
using all the data but including in the multiple
regression its equivalent variables that determine the
cycle. But we don't have such a theory. We know certain
things. We know that these cycles are irregular in
amplitude, they are irregular in timing, we know that we
dont have a satisfactory explanation. And given those
limitations of our knowledge, we want to suppress the
information about the cycle.
I wouldn't call it throwing away [information]. I
31
2.2.1 Phase Averaging
In illustrating the technique of phase averaging, we
follow standard F/S procedure and draw on NBER reference cycle
dates. Via this route, the phase average is computed as a
weighted average of all observation during a phase, an average
(or point) is obtained, and the procedure is repeated for
another phase. This is such that we get initial and terminal
points which demark a period for which a trend line may be
filtered.
In phase averaging, initial and terminal points are
weighted onehalf and intervening observations given a weight
of unity, as illustrated in appendix B. This procedure
constitutes imposing a prior belief on the data, which came
about from the use of the NBER chronology for the United
States and Economic Trends and Employment Gazette chronology
for the United Kingdom. The phase averaging helps in fitting
trend lines and thereby in separating the phenomena we wish to
investigate. It is also technically regarded as filtering
(Frazer 1988, 756; Jazwinski 1970).
In phase averaging and fitting trend lines, we separate
the components of information contained in a time series, and
as Friedman said, we highlight "one class of information" and
"avoid its being diluted by a class of information relevant to
would call it, rather... highlighting one class of
information and trying to avoid its being diluted by a
class of information relevant to a equation we're not
trying to answer. (Frazer and Sawyer 1984)
32
a question we're not trying to answer" (Frazer and Sawyer
1984). Indeed, the purpose of "phase averaging" is twofold—
to aid in the selection of the beginning and terminal points
for a period and to aid in fitting the trend line and
filtering out irrelevant information and keeping the more
permanent components. We may illustrate this use of a trend
line by refering back to figures iia and lib. Further, in
refer ing back to chapter 1 (sect. 1.2) we see the periods
themselves as episodes (i.e., periods charaterized by
different political/econoimic orientations) .
Also, F/S saw structural change— i.e., change in the
structure underlying formation of inflationary expectations.
And an episode entered i.e., the great peacetime inflation
with monetary accommodation of wages, government spending, and
prices, m addition, Frazer treated the turn in policy which
he referred to as "The Big UTurn." with that change we have
the monetarism that Thatcher and Reagan implemented in the
1980s. So with these two episodes we have the trend in the
monetary indicator for the 1970s, and the trend for the 1980s
and the turn from one period to the other for the United
States and United Kingdom, respectively. We see all this in
data results for both in the U.S. and the U. K. , which we
point to in figures lia and lib. The trend lines in the
figures are compatible with the economic/political orientation
we offer. They were obtained by the statistical method we
review in appendix C.
33
22.2 T he Filtering of Trend Li tips
There are several methods for filtering trend lines, (l)
the method F/S used; (2) fitting by a use of the simple
loglinear regression, and using beginning and terminal points
for a period such as may be obtained by the use of reference
cycle dates; (3) the equallength subsample method. in
actually obtaining trend lines in this present work, we use a
combination of the first and second methods, and H/E use the
third method. The results for both the U.S. and the U.K. sets
of data are reported in appendix C. The data we analyze are
the differences between the growth in wages and the growth in
productivity and the indicator of monetary policy both for the
U.S. and the U.K. The tests employed in finding the best fit
are two primary tests in an analysis of covariance. That will
test which model is the best fit (Frazer 1973, sect. 2A3).
The first of the primary test is a test of the hypothesis
of no interaction. in a regression context, where we treat
the categorical variables as control variables, the null
hypothesis of no interaction corresponds to the null
hypothesis that the N regression lines between Y and X for the
N levels of the categorical variables are parallel. If the
null hypothesis of no interaction is not rejected, then in
further analyses we assume that the N regression lines are
parallel.
The next hypothesis that is of interest is that the N
regression lines are in fact identical; that is, they have not
34
only have the same slope, but also the same Yintercept (see
appendix C) . The results obtained from illustrating the fit
of trend lines will be drawn upon in the following chapters,
when we test the hypotheses of section 1.1.2.
2  2  3 H/ E's EqualLength Subsamp lg.
In the F/SH/E controversy, a main question is whether
F/S's or H/E's results mean anything in terms of policy.
Drawing on one of H/E's latest articles (1990) on the issue of
modeling money demand, we find that they have attacked F/S's
use of "phase averaging" in the past, even as they themselves
engage in data transformations in terms of what they call the
"equallength subsample." m dealing with "equallength
subsample" H/E are taking time series and dividing them into
ten approximately equal length subsamples. m introducing
them H/E fit trend lines to the ten subsamples which they
illustrate and which we reproduce as figures 2la and 2lb.
The results they obtain by such an arbitrary division of the
time period give rise to their claim that "... virtually
every possible correlation between the growth rates of money
and prices can be observed" (H/E 1990, 12). These results
offers no positive policy associations, and, at the same time
they leave behind questions that we want to raise.
Whereas H/E pointed to "phase averaging" as losing
information in the earlier time (H/E 1983, 6) , they now use a
. 03 00
on
* * \
35
Figure 2la H/E's "EqualLength Subsample" for the Relation
between Money Stock and the Price Index for the
United States
36
. 09Q0
on
L
 . aiaa
*•% *
IV.
 . uiu .. uea . oia
oau . aaa . a4a
op
Figure 2lb H/E's "EqualLength Subsample" for the Relation
between the Money Stock and the Price Index
for the United Kingdom
37
method they refer to as an "equallength subsample." The
difference between this and "phase averaging" is that F/S
calculate phase averaging base on the chronology data provided
by the National Bureau of Economic Research and H/E pick up
their subsamples by focusing upon ambiguous, equally divided
subperiods where there is no economic reason for doing so.
Indeed Friedman considered that phase averaging separates
relevant information from information which may confound
estimation of the parameters. In this sense, Friedman's use
of phase averaging is likely to have a more stable relation
than for the original, unfiltered data, because the positive
serial correlation within a phase is at least partially
removed, and because the effects of extreme expansions and
contractions are dampened. The reverse of these reasons is
what leads H/E to pick up all the irrelevant information to
conclude with uncertain results and no positive policy
associations.
To summarize, phase averaging may not be the best filter
for the entire set of observations. Nonetheless, it is
justified as a mechanism for filtering out some episodic and
transitory changes and for thereby focusing on a more
persistent component of imformation.
2.2.4 H/E and Velocity as a Random Walk
In returning to the subject of velocity as a random walk,
Huhne in The Guardian said:
Professor Hendry likens velocity to the walk home
of a drunken man: he's heading roughly in the right
38
direction but one can never predict whether his
next step will be backwards, forwards or sideways.
It is a "random walk" (Huhne 1983)
While econometricans measure stability in terms of random
variability of the coefficients, F/S's conception of stability
is markedly different from that. For F/S, stability in the
velocity of money is stability as a phenomenon but not as a
numerical constant which means the explantory power and
magnitude of coefficients should not vary greatly across
episodes. Rather F/S make inferences about the phenomena and
the prospect that a given economic variate must be essentially
the same for two or more different economies (Frazer 1988 753
757) .
In 1990 H/E contradict what they claimed at the earlier
time. More recently they said, "As will be seen, the data for
both countries are remarkably similar in many respects,..."
(H/E 1990, 11). Here, indeed, the random walk hypothesis is
refuted by H/E's words about the close concordance of velocity
movements in the U.S. and the U.K. If velocity in each
country is indeed a random walk such parallel movements in
velocity will surely not be observed. That is one of the
reasons F/S introduced the extra dimension by way of the
consideration of two countries, rather than one.
2 . 3 The Multiple Regression Problems
Starting with Lawrence Klein, and following in most
econometric textbooks, we find a common emphasis on the
following: the determination of the values for the variables
39
within the models almost entirely; the multivariate regression
models; the avoidance of any role for episodic change (except
as may appear with the dummy variable) , and the assumption of
independence on the part of righthandside variables in
regression equations. Such an approach is symbolized by the
developments associated with big models and the Nobel
laureates Jan Tinberger and Lawrence Klein. The latter moved
from the Keynesian ISLM model to the Klein/Goldberger model
(Theil 1971, sect. 9.89.9) to the big models (Frazer 1984;
1991a, sect. 12. 4) .
In summary, the multiple regression problems we point to
are contrasted with the key features of Friedman's approach
(sect. 2.1) :
The Multiple
Problems
Regression
1. The determination of
value for the endogenous
variables within the model,
with a nonsignificant role
for the episodes.
2. The addition of variables
(also meaning time series)
to the righthand side of
the multiple regression
equation in an attempt to
account for unexplained
variation in a lefthand
side variable.
3 . The assumption of
independence in the
variables on the righthand
side of the multiple
regression.
Features
Approach
of
Friedman's
1. Attention to episodic
change, in combination with
a search for a stable
relation after adjustments
and allowences for episodes.
2. Friedman's decomposition
of time series with respect
to episodic and transitory
changes, and the more
permanent components.
3 . Interdependence in the
variables (also meaning time
series) due in large part to
episodic change.
40
4. The inability to set the 4. Attention to the bounds
bounds on the "true" on regression coefficients
regression coefficients as for the simple regression
variables are added to the model (i.e., also a
righthand side of the recognition of the "Learner
regression equation. problem") .
The features within each of the distinct columns are
interrelated. As to Friedman's use of statistical methods,
with the monetary policy emphasis he provides, Frazer reduces
the focus of the statistical work to one of the purpose at
hand and the usefulness of the method for that purpose (1991a,
sects. 2.4b, 3.2b, and 7.2c). In the monetary policy context,
we should treat effects along lines that are useful for those
who make and try to understand the monetary policy and effects
which comes about in a monetary economy.
On the one hand (the left col. above), there are the
efforts to seperate the effects by directing attention along
one line for the use of statistical methods, and, on the other
(the right col. above), there is virtual recognition that
success cannot be attained along the first line (hypothesis 4,
sect 1.1.2). To simply illustrate the line where success
cannot be attained, we have the following equation,
GNP =a + a,i + a 2 $/£ + a 3 P e + a 4 (funding policy) +... (2.1)
where GNP is gross national product, $/£ is the price of the
pound (£) expressed in U.S. dollars, i is the rate of interest
(symbolic of a vector of many rates) , "funding policy" is the
Bank of England's policy with respect to the government's
borrowing requirement and »...» signify omitted
41
variables. Now, the separation of effects problem is that
of thought and discourse where both proceed with the view that
the effect of righthand side variable (say, i) on the left
hand side variable (GNP) is separate from all the other
variables on the righthand side of the equation (say, $/L,
P e , funding policy and so on) , when in fact the effects of the
righthand side variables may be inseparable and indeed even
appear as a package of interdependent variables. For example,
inflation may be due to the government's "funding policy" (or
monetary accommodation of the government's financing) and at
the same time inflation may lead agents in the financial
markets to trade bonds at lower prices (higher interest rates)
and pounds at lower prices (lower exchange rates) .
The common, artificial means of proceeding is that of
"forcefully" locking up all other things explicitly or by
implication. The phrase " all other things" ( ceteris paribus
and presumed independence, and so on) may be invoked
explicitly, but most likely the presumption of independence
will be present only by implication.
Friedman recognized this separation of effects problem in
the analysis of actual data and sought to deal with it by
proceeding with what we have referred to as an alternative and
called the "indirect method" (Frazer 1988, 542) .
Starting with Walras, economists over the years added
features to the mathematical problem of having a solution to
an equation system when the number of variables equaled the
42
number of nonredundant equations. For one, the solution
equations came to be called reduced forms; for another, a
special class of parameters called exoqenous variables were
added; and for still another, distributed lag relations were
added, such as we have given attention to on one occasion or
another. The variables determined by the structural equations
models were said to be endogenous as taken up in appendix A.
The exogenous variables could be controlled, as by an outside
authority such as the Federal Reserve or the legislative and
executive branches of government. But the point was that for
the most part the parameters attached to the endogenous
variables of the model would be stable over time.
In statistical parlance, if the parameters had been
estimated for one period, then they will remain unchanged when
estimated for another sample period. 2 Unstable parameters,
as we are to see in the later chapters, mean that some
relevant real world forces were excluded from the regression
equation. Where the estimated values for parameters were not
stable and where the values for the error term are correlated
with the variables within the model, the idea in terms of
structural equation model thinking is to add more variables
and equations to account for the instability. The general
The sample period is the period over which data are
analyzed in the sense that they are used to estimate
parameters of a model. The model is presumed to apply
beyond the sample period in the classical, relative
frequency approach to probability. The parameters are
presumed to be stable in this approach.
43
idea, in other words, is to include all endogenous variables
in the model.
The problem arose in this approach that there was no end
to the number of variables and equations one could add
(Frazer 1973, chap. 14; 1984, 5153). We see Keynesian
economics move from a twoequation ISLM model, to a Federal
ReserveMIT model of the mid to late 1960s of from 65 to over
150 equations. By the early 1980s, Lawrence Klein had a model
with over 1000 equations and links to models for other
countries (Klein 1983) . Still there was no adequate stability
in the parameters that made them of any use for the purpose of
conducting monetary policy. Said differently, you still have
something very much like equation (2.1) above with the idea
that righthand side variables can be controlled independently
of one another.
Something was overlooked. On the one hand, it has been
suggested that instability had to do with random phenomenon
and, on the other hand, it has been suggested that omissions
had to do with psychological and socioeconomic forces, and
with learning, changing expectations, changing political
administrations (policy) , and so on. In summary, there are
two things: (1) there are impacts on the time series (such as
whether monetary accommodation or monetary discipline in the
U.S. or the privatization of government owned companies in the
U.K.) which interfere dramatically with the notion that the
inside variables are really determined within the analytical
44
system; and (2) there is the notion of upper and lower bounds
to a true regression coefficient (F/S 1982, 224226). The
latter calls attention to the frailty of the multiple
regression technique for analyzing economic time series, where
the purpose bears on the separation of the effects posited by
economic theory and the use of the equations as a guide to
economic policy.
F/S, in their book Monetary Trend (1982) , pointed out why
they proceeded differently. "We believe," they said, that
their indirect approach "yields insights that cannot be
obtained from the more sweeping approach (the pre1982
"fashion in econometric work")" (1982, 211). They believed
and said "that multiple correlations of many variables are
almost impossible to interpret correctly unless they are
backed by more intensive investigations of smaller sets of
variables" (1982, 214). F/S's indirect approach then is
simple and in contrast to proceeding immediately to compute
multiple regressions "including all variables that can
reasonably be regarded as relevant" (1982, 214).
The F/S use of simple method consisted of analyzing data
for a few variables at a time, before proceeding to the more
sweeping use of technique. Along this route, we encounter the
idea of a 'true 1 regression coefficient ( F/S 1982, 226) for
two variables and the extension of the idea by Edward Learner
to more than two variables (Learner 1978 and 1985; F/S 1982,
224225) . Starting with two variables from a purely
45
statistical point of view, there is the problem of all
variables being subject to error and the matter of setting
upper and lower bounds on the true regression coefficient.
Friedman said "that applying an upper and lower limit is
really the most effective way to have some idea of knowing
what I do know and what I don't know." Learner in turn replied
to a paper entitled "What Will Take The Con Out of
Econometrics ?" In this reply he discussed the extreme bounds
analysis (EBA) and the properties the bounds depend on (Learner
1985) . Continuing then, we have the Learner problem, which is
as variables are added to the regression equation "it's
extremely difficult to set limits on separate regression
coefficients . . . , and that beyond some points, you may be able
to set no bounds on it at all" (Learner 1985, 313).
CHAPTER 3
THE MONETARY INDICATOR, INFLATION RATES, AND MONEY DEMAND
3.1 Introduction
This chapter centers about hypotheses 1 and 2, which we
first stated in section 1.1.2. In the first of the hypotheses
we follow F/S in Monetary Trends (1982) . There they find that
the United Kingdom and the United States have in common the
same determinants of the money demand functions (F/S 1982,
sect. 5.4). Others such as H/E argue that velocity is a
random walk (sect. 2.2.4), yet statistical results indicate
that velocity cannot be a random walk. If velocity in each
country is indeed a random walk, then the parallel movements
that we find in velocity will surely not be observed.
In the second hypothesis we take up the question whether
prices and nominal wage rates adjust more readily in the
presence of monetary discipline as defined in section 1.2. We
recall that Keynes pointed to "sticky wage" as a behavioral
matter in the 192 0s and based the General Theory on the notion
of a wage standard as reviewed by Hicks (1983) — i.e., that
wages would remain "sticky" and hence vary only in relation to
productivity, as total spending was managed to achieve
Keynesian full employment.
Hypothesis 2 gains its importance for having been
introduced by J. M. Keynes in the 1920s, in the form of the
46
47
issue of sticky prices and the special price called "the wage
rate." As built into Keynes's General Theory f!93 6^ and as
introduced in section 1.1.2, wages failed to adjust downward
in the presence of unemployment, there was a failed view of
monetary policy, and fiscal policy enters.
As reviewed by Frazer (1988, chap. 16; 1991b), Friedman
offers a substitute view to that of interest rates and
monetary policy, and it was taken up by Ronald Reagan in the
United States as a part of what was called "supplyside
economics." The idea was to use monetary policy to tame
inflation and hence in general bring about a downward
adjustment in the inflation rate which had gotten built into
pricing policy and wage contracts in the 1970s. In the United
Kingdom Margaret Thatcher also adopted Friedman ' s "monetarism"
(so called in London) and took up the issue of wage
adjustments as a part of the taming of inflation by monetary
means. via this route we link up the economics and the
political positions and point to what we have defined in
sections 1.1.1 and 1.2 as monetary discipline.
In the context of the foregoing analysis, an empirical
question arises. That is whether the inflation is caused by
nonmonetary forces (power theories of inflation) or by
monetary forces (accelerated money growth). 1
Power theories of inflation center about the economic
theories of market structures (Frazer 1991a, sect. 2.2e
and 2.2g). Early on Friedman tended to reject them
(Frazer 1988, 306323).
48
3  2 The Monetary Indicator. M oney Demand: an Episodic View
In returning to hypothesis 1 (sect. 1.1.2), figures lla,
1lb, may be referred to again as well as the fitting
procedure we take up in appendix C. We recalled that the
figures showed the trend in monetary policy for the 1970s
(actually 1970:1 to 1979: IV), the 1980s (1980:1 to 1989:11),
and the turn from one period to the other for the United
States and United Kingdom respectively. To indicate the
periods and the turn, we use the indicator provided by the
product of the change in policy [A (1/M) (dM/dt) ] and the
velocity ratio (Y/M) . The velocity ratio enters as a factor
on two grounds: (1) because it is a measure of the impact of
the policy change on total spending, and (2) because monetary
officials should be expected to take the secular shifts in the
velocity ratio into account in judging the impact of the
policy they pursue.
In reference to the trend lines (figures lla and 1lb)
the equations we obtained are as follows:
M~= 2.440101 + 0.2772t ( during 1970s )
(8.456)* (2.9785)*
M~= 2.558236  0.1794t ( during 1980s )
(9.225)* (2.6246)*
for the United States, and
M~= 1.599123 + 0.1776t ( during 1970s )
(4.1175)* (2.6016)*
49
M~= 22.53880  1.8899t ( during 1980s )
(8.9241)* (4.5112)*
for the United Kingdom. In these equations, t is time, M~ is
the indicator of monetary policy (expressed as percentage
point) , the asterisk means coefficients are significantly
different from zero at 5% level of significance.
Comparing the equations and the related results for the
1970s with those for the 1980s, we see a significant
difference between the two decades with respect to the
indicators of monetary policy. During the 1970s the trend was
upward in the U.S. by almost 0.27 percent per annum. Yet,
turning into the 1980s the trend was downward by almost 0.18
percent per annum. In the United Kingdom, during the 197 0s
the trend was upward by almost 0.18 percent per annum. Yet,
turning into the 1980s the trend was downward almost 1.9
percent per annum.
Upon comparisons, two major points are possible. First,
the trend lines for the United States and the United Kingdom
reveal a significant difference between the two decades.
Second, both U.S. and U.K. trend lines are moving upward in
the 1970s and downward in the 1980s. With the governments
adopting a full employment goal, without a major regard for
the inflationary consequences of monetary policy, they
accommodated other developments such as wage price increases.
Under monetary discipline, such as we find for the 1980s for
the most part, the wage and inflation rates may be expected to
50
adjust downward to achieve employment up to the natural rate
(i.e., to achieve the noninf lationay rate of unemployment).
Returning to the hypotheses, the trends we reported for
the monetary policy are found in both the U.S. and the U.K.
Said differently, the respective countries shared common
monetary policies. The results from the analyses of data for
the 1970s and 1980s decades support F/S's claim in Monetary
Trends (1982, sect. 5.4).
Also, with respect to the results and the distinction we
make between the 1970s as an episode and the 1980s as an
episode (sect. 1.2), we see a case for the importance of the
episodic approach to data analysis. Certainly the case seems
meaningful by comparison with H/E's case of the equallength
subsample (sect. 2.2.3). Also from these results we see that
the series change direction from one episode to another
episode.
The further implication of the results is that the same
causal, exogenous forces affected the series in the two
countries, namely, monetary policy. So, in addition to
examine the indicator of monetary policy, we move one more
step to see whether the series for money demand exhibit
similar results, as suggested by F/S's Monetary Trends (1982,
sect. 5.4) and hypothesis 1 (sect. 1.1.2).
Fitting trend lines for various income velocity ratios,
we obtain the following results, shown as figure 3 la and
figure 3lb:
51
VI = 4.490945 + 0.199708t ( during 1970s )
(126.517)* (35.121)*
VI = 8.193847  0.097976t ( during 1980s )
(35.677)* (6.767)*
for the United States, and
VI = 5.249726 + 0.166599t ( during 1970s )
(52.765)* (9.486)*
V3  5.394873 + 0.153929t ( during 1970s )
(27.962)* (4.526)*
VI = 11.94126  0.390892t ( during 1980s )
(42.160)* (20.535)*
V3 = 10.363697 0.308630t ( during 1980s )
(42.160)* (20.535)*
for the United Kingdom. In these equations, t is time, VI is
the velocity ratio for dollars Ml or pounds Ml as the case may
be, V3 is for pounds sterling M3 in the United Kingdom. And
the asterick means that the coefficients are significantly
different from zero at 5 percent level, in these statistical
results we again see a significant difference between the
orientations of the 1970s and the 1980s respectively. In the
case of U.S., during the 1970s the trend was upward by almost
0.2 every year as both shown in the statistical results and as
illustrated in figure 3la. Yet, turning into the 1980s the
trend moved downward by almost 0.1 every year as was also
shown and illustrated. Roughly paralleling these results for
52
8
7
6
5
4
3
ml velocity
H 1 1 1 1 1 1 1 1 1 h
H 1 1 h— * 1 1 1 1 1 h
1968 1972 1976 1980 1984 1986 1992
Year
Figure 3la Money Demand in the U.S. during the 1970s and
1980s Respectively
 GNP/M1
GNP/M3
53
8
7
6
5
4
3
H 1 1 1 1 h
H 1 1 1 1 1 1 1 1 1 1 1 1 1 h
1968 1972 1976 1980 1984 1988 1992
Year
Figure 3lb Money Demand in the U.K. during the 1970s and
1980s Respectively
54
the U.S. and illustrated in the figure 3lb the trend in the
U.K. was upward by almost 0.15 every year during the 1970s
both in the series VI and series V3 . Yet, turning into the
1980s the trend moved downward by almost 0.3 every year for
series V3 and 0.4 for series VI.
These foregoing results again show that both for the U.S.
and the U.K. the series moved in the same direction in the
respective decades. Consequently, the results show further
support for the prospect that income velocity is not moving as
a random walk. If money demand did behave as a random walk,
then parallel movements in the series of income velocity would
not be found.
Making the twocountry comparisons introduced by F/S
(1982), we see not only parallel movements the U.S. and the
U.K. series but a shared change of direction from the 1970s as
an episode to the 1980s as an episode. The implication is
that the same causal forces affected the series in the two
countries.
The analysis we presented with above and the results we
found are supportive of hypothesis 1 (sect. 1.1.2). The two
countries are experiencing similar monetary phenomena.
3  3 Prices, Nominal Wage Rates and Monetary Discip line
Returning to hypothesis 2 (sect. 1.1.2), we recall that
Keynes did two things which are of present interest, namely,
(1) Keynes pointed to "sticky prices" (and the wages
underlying them) in the 192 0s, and (2) he based the General
55
Theory on the notion of a wage standard (Hicks 1974; 1984).
The first of these notions is the one primarily addressed by
hypothesis 2, and the second is closely related. In the first
case, we simply compare the inflation rate adjustment for the
1970s and 1980s respectively, where in the first decade we
have monetary accommodation and in the second decade we have
monetary discipline. We expect, in contrast to the "sticky
prices" view, that the inflation rate will adjust downward in
the 1980s, in both the U.S. and the U.K., plus we expect
greater adjustment in the U.K. than in the U.S., principally
because the discipline was greater in the U.K. (figures 1ia
and 1lb) .
The second notion above—that of the wage standard
further implies that wages will not rise greater than
productivity (i.e., that wages and productivity bear a
constant and unchanging relation to one another that neither
is altered by inflation or deflation) in the presence of
demand management to assure full employment. The second
notion then is also related to the first to the extent that
the use of monetary policy (or demand management policy
generally) should have no bearing on the inflation rate either
in terms of the price indexes or the relation between nominal
wages and productivity.
To test hypothesis 2 we presently do two main things: (l)
follow Friedman and examine relations between trends in the
data series, and (2) split the overall sample period into two
56
subsample periods. In the latter case, we also follow the
view that monetary policy is very different in the respective
subsample periods. Although we look at the secular trend
data and the policy differences in the two decades, there is
the prospect, from the policy point of view, that inflation
rates and nominal wage rates in relation to productivity
respond differently to monetary accommodation, on the one
hand, and discipline, on the other. We take this to mean that
monetary discipline facilitates price and wage adjustments.
The reliance on the trends in the data series, as we
pointed out initially (sect. 2.2) is simply a way of filtering
out of the time series some of the troublesome, episodically
imposed information we do not wish to confront. The data
method we employed in this regard may be readily contrasted
with that used by H/E (sect. 2.2.2).
In proceeding we obtained results of two types. First,
we juxtapose trends in the series for inflation rates, nominal
wage rate and the indicator of monetary policy for the
respective decades. Second, we examine means and standard
deviations for the inflation rate and the indicator of
monetary policy measures for the respective decades. The
differences between our analysis and that advocated by H/E are
twofold, that we give emphasis to knowing something about the
policy experiments that generated the data (as opposed to say
arbitrarily selecting the sample period) in the choice of the
sample periods; and that we readily admit to sampling from
57
different universes where H/E do not. In the one instance,
the emphasis comes from an exogenous, policy view of time
series. in the other, the emphasis comes from the "the
prevailing fashion in econometric work" where we encounter the
prospect sampling from the same universe.
3.4 Statistical Results
In this section we present the statistical results for
hypothesis 2. in the testing, we first look at trends
(figures 1la, 1lb, 32a and 32b) and related results. We
next look at the averages for the split sample period (figures
34a and 34b) and the related results.
3.4.1 The Trend Analy sis
The earlier figures 1la and 1lb may be referred to
along with the present figures 3 2a and 3 2b. Where we show
trend lines for the inflation rates for the United States and
the United Kingdom, respectively. The equations defining
these trend lines are as follows:
P cpi " 4.347273 + 0.780606t ( during 1970s )
(3.039)* (2.914)*
P gnp = 5.976364 + 0.316364t ( during 1970s )
(6.967)* (2.699)*
P cpi =10.845968  0.437903t ( during 1980s )
(2.889)* (2.607)*
P GNP =11.548065  0.504194t ( during 1980s )
(3.639)* (2.285)*
for the United States, and
58
P CPI = 10.874545 + 0.647879t ( during 1970s )
(3.473)* (2.104)*
P GNP = 8.930909 + 0.870909t ( during 1970s )
(3.162)* (2.386)*
P CPI =12.043387  0.410161t ( during 1980s )
(2.684)* (2.314)*
P GNp =24.115484  1.166452t ( during 1980s )
(3.639)* (2.485)*
for the United Kingdom. In these eguations, t is time, P cpj is
the inflation rate CPI, P GNp is the inflation rate measured by
the GNP deflator. The asterisk means the coefficients are
significantly different from zero at 5% level.
In response to the accommodative monetary policy in the
1970s, we observe positive slope for the trend lines
of the inflation rates both in the United States and the
United Kingdom. Also, for the 1980s era of monetary
discipline, all the trend lines for the inflation rate series
take on negative slopes not only for the United States data
but also for the United Kingdom data.
The above results indicate very positive support for
hypothesis 2~prices adjust more readily in the presence of
monetary discipline. Indeed, inflation rates readily adjust
downward in the presence of monetary discipline both the U.S
and the U.K. .
Taking up the same procedure as that used above, and
extending its use to the nominal wage rates for both the U.S.
59
inflation rate (deflator of GNP)
inflation rate (CPI)
1968 1972 1976 1980 1984 1988 1992
Year
Figure 32a Trend Lines for Inflation Rates, United States
60
 inflation rate (deflator of GNP)
— inflation rate (CPI)
1968 1972 1976 I960 1984 1988 1992
Year
Figure 32b Trend Lines for Inflation Rates, United Kingdom
61
and the U.K., we obtain the following trend lines:
w = 1.7064 + 0.0613t ( during 1970s )
(5.212)* (3.909)*
w = 3.8640  0.l670t ( during 1980s )
(7.342)* (7.227)*
for the United States, and
w = 3.1917 + 0.0745t ( during 1970s )
(4.572)* (1.134)
w = 5.3887  0.1974t ( during 1980s )
(9.381)* (5.167)*
for the United Kingdom. in these eguations, t is time, w
is the nominal wage rate (hourly compensation) . The asterisk
means the coefficients are significantly different from zero
at 5% level. The results are illustrated graphically in
figures 33a and 33b.
Again, viewing the results overall, they indicate stong
positive support for the wage part of hypothesis 2— nominal
wage rates adjust more readily in the presence of monetary
discipline. In response to the accommodative monetary policy
in the 1970s, we observed positive slopes for the trend lines
for nominal wage rates both in the United States and the
United Kingdom, and we also observe that during the monetary
discipline era of the 1980s. All trend lines for the series
of nominal wage rates take on negative slopes for both the
United States and in the United Kingdom.
62
%
10
8
f— *
o
o
H
*
* 6
■n
* 4
2
H 1 h
h — i — i — h — i — i — i
1968 1972 1976
• — i — i — i — i — i
1980
YEAE
H 1
1984 1988 1992
Figure 33a Trend Line for the Nominal Wage Rate,
United States
63
o
o
%
10
8
6
? 4
;:/\
H 1 1 1 1 H
H 1 1 1 h
♦ — I — I — 1 — I — I — I
1968 1972 1976 1980 1984 1988 ' £ 92
YEAE
Figure 33b Trend Line for the Nominal Wage Rate,
United Kingdom
64
This found shows nominal wage rates adjust readily in the
presence of monetary discipline. In addition, the nominal
wage rates for both the U.S and the U.K. reflect to the change
in monetary policy as we move from the decade of the 1970s to
that of the 1980s.
In order to be reassured of the relationships between the
prices, the nominal wage rates, and the monetary policy, we
move one step further to examine relations between the
predicted inflation rates and the predicted indicator of
monetary policy. m doing this we obtain the following
equations:
p C Pi e = 15.36301 + 2.816039M~ e ( during 1970s )
(13.55)* (3.994)* R 2 = .98
P GNP e  6.268468 + 1.141284M~ e ( during 1970s )
(18.67)* (4.599)* R 2 = .99
P cPi e = 8.57694 + 2.440931M~ e ( during 1980s )
(39.6)* (15.94)* R 2 = 0.99
P G N p e " 988474 + 2.810446M~ e ( during 1980s )
(39.9)* (16.56)* R 2 = 0.99
for the United States, and
P cpi S = 5.041080 + 3.647967M~ e ( during 1970s )
(6.84)* (7.554)* R 2 = .99
P GNP 6 " 1.089295 + 4.903766M~ e ( during 1970s )
(4.237)* (6.321)* R 2 . .99
p C Pi e = 0.653348 + 2 . 170278M~ e ( during 1980s )
(3.595)* (5.94)* R 2 . .99
65
P GNP 6 " 0.844593 + 6.172630M~ e ( during 1980s )
(4.114)* (10.22)* R 2 = 0.98
for the United Kingdom. In these equations, M~ e is the
predicted indicator of monetary policy, P cpi e is the predicted
inflation rate (CPI) , P GNp e is the predicted inflation rate
(deflator of GNP) , and the asterisk means coefficients are
significantly different from zero at 1% level. Again, to be
sure, the results we have here indicate that the predicted
inflation rates (ex post predictions) are highly correlated
with the predicted indicators of monetary policy (also ex
post) in both the United States and the United Kingdom.
Similar results are obtained for the relations between
the predicted nominal wage rates and the predicted indicators
of monetary policy. We obtain equations as follow:
w e = 1.220088 + 0.221140M~ e ( during 1970s )
(7.865)* (3.514)* R 2 = 0.99
w e = 3.273653 + 0.930881M~ e ( during 1980s )
(39.9)* (10.08)* R 2 = 0.99
for the United States, and
w e = 0.025209 + 0.419482M~ e ( during 1970s )
(3.145)* (6.334)* R 2 = .99
w e = 0.273694 + 0.104450M~ e ( during 1980s )
(5.266)* (4.411)* R 2 = 0.99
for the United Kingdom. In these equations, M~ e is the
predicted indicator of monetary policy, w e is the predicted
nominal wage rate, and the asterisk means coefficients are
66
significantly different from zero at 1% level. The results we
have here indicate that the predicted nominal wage rates have
positively correlated with the predicted indicator of monetary
policy in both the United States and the United Kingdom for
the different decades.
In summary, we have the predicted monetary policy
controlling the predicted inflation rates and the predicted
nominal wage rates for the 1970s and 1980s in the U.S. and the
U.K. .
3.4.2 The Mean Analy sis
As we may see from the trend analyses of both the U.S.
and the U.K. data in the previous section, the respective
countries shared an accommodative monetary policy in the
1970s, in a move toward monetary discipline in the 1980s, and
in the taming of inflation in the 1980s. To consider those
changes further, we presently examine the averages for
monetary policy indicator and the inflation rate data.
In table 31 we show the maximum, minimum, mean, and
standard deviation for the indicator of monetary policy for
both the U.S. and the U.K. . As for the previous analyses, the
overall sample period is split into the two subsample
periods. m addition to table 31, we illustrate the
statistical results graphically in figures 34a and 34b.
Whereas the mean for indicator of monetary policy for the
United States in the 1970s is positive (3.388 percent), for
the 1980s the mean of indicator of monetary policy takes on a
67
Table 31 MAXIMUM, MINIMUM, MEAN AND STANDARD DEVIATION FOR
THE INDICATOR OF MONETARY POLICY
Periods of time
Min.
Max.
Mean
S.D.
United States
1970s (iregime)
1980s (Mregime)
0.258%
10.98%
7.463%
7.987%
3.388%*
0.267%*
1
4
.760%
.800%
United Kingdom
1970s (iregime)
1980s (Mregime)
5.766%
16.506%
13.477%
13.188%
2.642%*
4.141%*
5
7
290%
011%
The asterisk means of the indicator of monetary policy
are significantly different toward one another at 1%
level in each country 2
negative sign (0.267 percent). Moreover, similar results
appear for the United Kingdom. For the U.K. the mean of
monetary policy indicator is 2.6421 percent in the 1970s, and
the mean of the indicator of monetary policy becomes 4.141
percent for the 1980s. Comparing the U.S. and the U.K. data
on the indicator, we see that the U.S. appears somewhat less
resolute in taming inflation for a time following the 198182
recession and at the close of the 1980s.
Against the background of the indicator results and
hypothesis 2 (sect. 1.1.2), we should observe a move from the
high inflation rates of the 1970s to lower inflation rates in
the 1980s. Infact, the maximum, minimum, mean, and standard
 The tests of the significance of the difference between the
mean of the indicator of monetary policy are shown in
appendix E.
68
Table 32 MAXIMUM, MINIMUM, MEAN AND STANDARD DEVIATION FOR
THE INFLATION RATES
Periods of time
Min.
Max.
Mean
S.D.
United States
1970s (iregime)
deflator of GNP
4.7%
9.8%
7.4%*
1.676%
CPI
3.2%
13.5%
7.86%*
3.293%
1980s (Mregime)
deflator of GNP
2.5%
9.7%
4.43%*
2.263%
CPI
2.0%
10.3%
4.66%*
2.4%
United Kingdom
1970s (iregime)
deflator of GNP
CPI
1980s (Mregime)
deflator of GNP
CPI
The asterisk means of the inflation rates are
significantly different toward one another at 1% level in
7.0%
7.0%
27.2%
27.4%
12.85%*
13.79%*
5.992%
5.392%
3.5%
3.5%
19.5%
11.9%
7.65%*
6.25%*
4.999%
2.705%
each country
deviation for the inflation rates in terms of both CPI and the
deflator data for both the U.S. and the U.K. support the
hypothesis. in table 32, the means for the U.S. inflation
rate data are 7.4 percent for the deflator and 7.8 6 percent
for CPI per annum in the 1970s and considerably less in the
1980s (4.4 percent for the deflator of GNP and 4 . 6 percent for
CPI respectively) . Similar results were found in the case of
the United Kingdom, for the U.K. the means for inflation rates
are 12.85 percent for the deflator and 13.79 percent for CPI
per annum during the period of monetary accommodation, and
moving into the 1980s the means are only about half of those
in the accommodative period of time.
69
%
20
10
Mean of indicator of monetary policy in 70s
Mean of indicator of monetary policy in 80s
HtkflJtl»»
Wt>'
IJT
10
If
L
•20 H — i — i — i — l — i — i — i —  — i — i — i —  — i — i — i —  — i — i — i —  — i — i — i —  — i —
1968 1972 1976 1980 1984 1988 1992
Year
Figure 34a The Mean for the Indicator of Monetary Policy
1970s vs 1980s, United States
70
%
20
10 
10
Mean of indicator of monetary policy in 70s
Mean of indicator of monetary policy in 80s
20
1968
1992
Figure 34b The Mean for the Indicator of Monetary Policy
1970s vs 1980s, United Kingdom
71
Thus, viewing the mean analysis results overall in tables
31 and 32, the results support hypothesis 2, prices adjust
more readily in the presence of monetary discipline. In
summary, we see that the indicators of monetary policy moved
from positive sign (accommodative) into negative sign
(discipline) both in the U.S and the U.K., and that we move
from higher inflation rates to a lower inflation rates.
3.5 Summary
Based on the basis of results presented in chapter 3, we
find support for hypothesis 1 (sect. 1.1.2). Not only the
indicators of monetary policy move parallel in the respective
countries (figures 1la and 1lb) , the income velocities also
move parallel (figure 3 la and 3 lb) . Moreover, these shared
results indicate that the demand for money (i.e., the velocity
of income) cannot be a random walk as some have maintained.
If indeed velocity in each country is a random walk, then the
parallel movements we report for velocity will surely not be
observed. In setting the new results in opposition to H/E's
we support F/S's claim and reject H/E.
Drawing on the indicator and inflation rate results, and
others we also find support for hypothesis 2, namely, prices,
nominal wage rates adjust more readily in the presence of
monetary discipline. Extending the analysis to the nominal
wage rates, from figures 1la, 1lb, 32a, 32b, 33a and 33b
,not by coincidental, we find that reflecting to the
accommodative monetary policy in the 1970s we observed all
72
positive trend slope both in U.S. and U.K. based on two
different prices series for each country, also we find that
the series for nominal wages have positively sloping trends in
the 1970s and negative trends in the 1980s in both U.S. and
U.K. Further, we report that the predicted inflation rates
and the predicted nominal wage rates are highly correlated
with the predicted indicator of monetary policy. This finding
shows that prices and nominal wage rates respond to the
changes in the policy indicator. Said differently, monetary
policy is controlling the series of prices and nominal wage
rates. In the presence of monetary discipline we see prices
and nominal wage rates adjust readily.
In addition, we also examined means for the U.S. and the
U.K. prices. Via this route, we saw that the means for the
prices reflected the mean for the indicator of monetary policy
both in the United States and in the United Kingdom. While
the indicators of monetary policy move from positive sign
(accommodative) to a negative sign (discipline) both in the
U.S. and the U.K., the observed prices also responded and
moved from high inflation rates to lower inflation rates.
CHAPTER 4
MONETARY POLICY, WAGE, AND PRODUCTIVITY
4 . 1 Introduction
Hypothesis 3 states that wage rates are determined by
productivity and market structures irrespective of monetary
policy. This we associate with a theorem due to Sidney
Weintraub and Stephen Rousseas (Weintraub 1978, 283 0; chaps.
78; Rousseas 1986, 7477). It is presently viewed as a part
of the Keynesian/postKeynesian analytical system. In Lord
Kaldor's terms, the money stock is endogenous to the
analytical system. In this view, the explanation for
inflation resides in theories of market power, and no
attention is given to monetary matters except that fiscal
policy is to be accommodated. Here once again we encounter
the distinction between the decades of the 1970s and 1980s,
along the lines of Keynesian monetary accommodation, on the
one hand, and the alternative of monetary discipline to tame
inflation, on the other.
Once again in the testing of the hypothesis we also take
up the trend paths for the time series at issue and the mean
values for the respective decades. This treatment of the data
series parallels that of chapter 3.
73
74
4.1.1 The Post Keynesian Theory of Inflation
Keynes ians from J.K. Galbrath onward have been shown to
find the cause of inflation in terms of theories of market
power (Frazer 1988, 194208) rather than monetary forces such
as M. Friedman emphasized — even as Friedman rejected theories
of market structures other than those for the perfect market
and monopoly (Frazer 1988, 306323) . For the present,
however, we will focus on the post Keynesian view associated
with Weintraub and Rousseas, and juxtapose it with a view we
find in the works of Friedman and Frazer. The particular post
Keynesians we point to emphasize "oligopoly market power"
which permits a markup in prices over an historically set
wage level. This wage level they see as "the result of the
struggle between capital and organized labor over relative
shares" (Rousseas 1986, 76). Also in order to put monetary
policy into a "sustaining" role the post Keynesians claim that
prices are a function of nominal wages, and the wages are
exogenously determined by the process of collective
bargaining. They view wage compensation as being price
neutral (i.e., that wages change prices, but prices do not
change wages) . 1
In addition to that view, compensation in terms of money
wages takes place according to the average rate of
1 This neutrality position is the way the post Keynesian
proceed in ruling out the prospect for monetary
influences on wage. In order for monetary policy to be
endogenous in the post Keynesian view they see it as
accommodating to fiscal policy to assure full employment.
75
productivity. Compensation, consequently, is neutral with
respect to both the price level and labor's share of income,
ceteris paribus. But as post Keynesians raise the issue they
must deal with a one sector model, because in a multisectoral
model their contention does not hold. We offer a proof in
appendix D.
4.1.2 The RousseasWeintraub Theorem and Monetary Policy
Along the foregoing lines, we consider the wage theorem
which Rousseas finds in Sidney Weintraub's work and calls a
major tenet of American post Keynesian economics (Rousseas
1986 7377; Weintraub 1973, 2830). We call it the Rousseas
Weintraub wage theorem. It is "that, on the whole, prices are
determined by some markup over unit labor costs (Rousseas
1986, 74).", namely:
P  k (W/Q)
= k (W/L)/(Q/L) (4.1)
Here k is the given degree of monopoly in the economy. It is
determined by the exogenous, institutional environment within
which each firm operates. The W/Q factor is the ratio of the
total nominal wage bill (W) to the level of real output (Q) ,
plus it is a measure of the unit labor cost of producing that
the total output. By dividing the total nominal wage bill (W)
to the level of real output (Q) by the total labor input (L) ,
we see that the equation becomes:
P = k(W/L)/(Q/L) (4.1)
or
76
P = k(w/q) (4.2)
In this equation 4.2, w is the average annual wage rate in
nominal terms, and q the average productivity of employed
labor. It is assumed to grow at a relatively constant rate
over time.
Continuing Rousseas says, "if the relative increase in
the nominal wage rate exceeds that of the average productivity
of labor ( w°>q°), prices will rise (Rousseas 1986, 74)." He
writes,
P = P(w) (4. 3)
In this case, w is exogenously determined by the process of
collective bargaining. "In short," Rousseas says, "prices are
a function of nominal wages (Rousseas 1986, 74)," and the two
are positively related. Thus we have a major tenet of
American post Keynesian economics, which is "prices are a
function of nominal wage (Rousseas 1986, 75)."
Via this route, the post Keynesians attempt to nullify
the role money plays in determination of the price level. In
doing so they put power theories of inflation central to their
view of inflation. Rousseas says:
Essentially, . . . , as long as money wages
are exogenously determined around the
bargaining table, monetary policy has
only an indirect link to the price level.
(Rousseas 1986, 77)
Further, the post Keynesians put monetary policy into only a
"sustaining" role. Rousseas says:
The increase in nominal income, due to a
rise in unit labor costs results in an
77
increased transactions demand for money
for any given level of real output.
Therefore, if [in order ] real output and
employment are to be maintained, the
supply of money will have to increase
(Rousseas 1986, 75)
This position is very much that of Lord Kaldor, as taken up by
Frazer (Frazer 1988, 9798, 545, 740; and 1991a, sects. 3.5c
3.5e). Continuing Rousseas says, "If, as Weintraub assumes,
the velocity of circulation is constant, a full accommodation
will be required." (Rousseas 1986, 75) He then says:
If the central bank flatly refuses to
increase the money supply, then the
resulting excess demand for money will
cause interest rates to rise with the
expected Keynesian result of a fall in
investment leading to a decrease in real
output and employment... (Rousseas 1986,
75)
Returning to the connection with Kaldor, we find him saying
the following:
At any time, or at all times, the money
stock will be determined by demand, and
the rate of interest determined by the
central bank. (Kaldor 1982, 24)
To argue the endogeneity of money supply further, Kaldor says
the monetary authorities have no choice but to accommodate the
"needs of trade." He says:
The central bank cannot refuse the
discounting of 'eligible bills' rendered
to it Precisely because the monetary
authorities afford the disastrous
consequences of a collapse of the banking
system... the 'money supply' in a credit
money economy is endogenous, not
exogenous— it varies in direct response
to changes in the public 'demand' to hold
cash and bank deposits and not
78
1982 Pe 4?r tlY ° f that demand  ( Kald °r
In summary, the post Keynesian inflation theory may be
stated as follow. First, the government starts deficit
spending to achieve full employment, which in turn would
depends on credit expansion. Second, with the central bank
keeping open the discount window on an unlimited basis and
fully meeting support for credit expansion, the money supply
accelerates. However, for the post Keynesians this policy of
accommodation has no bearing on cost pressures which may get
push forward.
4,2 frazer /Friedman Wa ge Bargaining Therrry
In this section, we take up Friedman's early work on the
relationship between the wage rate and the expected prices
level, which Frazer embellished. In Frazer's analysis
monetary policy not only plays role in influencing price level
but it also influences the wage rates through the wage bargain
process and the presence of price indexes in labor contracts
( Frazer 1991a, sect. 12.3).
4 2.1 Frazer's Analy sis
Frazer offers a monetary approach to the analysis of
price averages, wage adjustments, and production (1980, sect.
17.2 and 1991a, sect. 12 .3). In it he achieves a
compatitability with Friedman's long run view of the Phillips
curve (Friedman 1982 sect. 12.2; Frazer 1991a, sect. 12.2).
The monetary analysis is in a dynamic context where
production (Q) is moving along a trend path at full employment
79
(Q f , say production at the "natural" or noninf lationary rate
of employment). Indeed, the actual production (Q) may be
viewed in relation to this trend path(i.e., Q/Q f ) such that
the actual rate varies about the full employment rate (namely,
! QQ f ! )
As an illustration, Frazer offers what we show as figure
41. There a price index appears on the vertical axis and the
ratio of output to full employment output appears on the
horizontal axis. On the plane determined by these axes,
supply and demand curves are imposed, by analogy to Marshall s
cross. The demand curve slopes downward, and the supply
curve, as shown, is kinked at full employment output denoted
by the Q/ Qf ratio where actual output equals full employment
output (Q= Qf or Q/Q f *i00 = 100 percent). As with Marshall's
cross, costs underlay the derivation of the supply curve, plus
labor costs (wages) are the major component of costs. There
are labor unions and production is accounted for mainly by
large manufacturing firms, in addition, the unions have cost
ofliving classes in the wage contracts and relate the reality
and prospect of inflation to higher nominal wages.
By the same taken Frazer relates the prospect of
inflation and wage adjustments to monetary accommodation and
discipline such as we defined in section l.i.i. A n
illustration of the analysis in reference to figure 41 starts
with inflation in progress at point A, which occurs at over
full employment (Q > Q f ) . Under such occurrence there is
80
Ratio of Output to Ful
Employment Output
Figure 41 The PriceOutputWages Connection
Aggregate Demand and Aggregate Supply
Source: Frazer Alternative Analytical System 1991a, 354
81
little discipline on the wage bargain between the two sides of
industry and wages are pushed higher (even in excess of full
employment wage rates) . Acting to assure "full employment" at
any inflation rate, the monetary authority accommodates
inflationary wage increases via the management of aggregate
demand. The demand curve shifts (say, from D t D, to Dp 2 ) , and
higher wages and inflation rates are accommodated. The issue
of accommodation and "sticky wages" (as oppose to wage
adjustments) is enjoined such as was set on a causes by J. M.
Keynes and Winston Churchill in the 1920s (Keynes 1925; Frazer
1988, 419421) .
The issue is later joined by Margaret Thatcher (Frazer
1988, chap. 15 and 1991a, sect. 12.3). The alternative she
poses is wage adjustment. Thatcher said:
Supposing we start off with inflation. You have it
We have it, at very high rates. Rates that have
gone up over the last decade to far higher rates
than we would have thought possible. And you also
do %L C % ^ am ° Unt ° f unem P lo ^ent. Now you can
s?ick?na fn^V Y °\ Can reflate. That means
w rtSi ;° n ° n t0p ° f infl ation, and what I
tZ ii l r SU , 1 , tCaSe mone y" Germany had it after
the first World War. When you get that you get
unemployment on a colossal scale. Now what°s the
alternative policy? You've got inflation. You try
dowrf Th e ^ POllCleS . that Wil1 * et the inflation
down That means not having so much surplus money
in the economy so that prices come down. Unless
amoSn? COndltion their wage claims to the lesser
amount of money, then there'll be some
s?m wan?^ Tv at US . Ua11 ^ happenS 1S that People
still want to take out quite a lot for themselves,
in uni^r ^^ ° Ver f ° r ° thers ' and ifc c ™es out
in unemployment. But in the longer run, you'll not
get a competitive industry, good secure jobs unlSss
That ZT S r^L COm P. etiti ve with other peoples?
That means fighting inflation now, it means short
82
run unemployment, but long term good jobs, good
prosperity, good prospects. (Frazer 1988, 611)
The alternative Thatcher posed in relation to the monetary
accommodation was monetary discipline which we introduced in
section 1.1.1.
Now, the matter we address for the 1980s need not be
viewed differently from that engaged in by Reagan's Presidency
or Thatcher's government. Moreover, the effectiveness of the
policy need not be viewed independently of Reagan's
confrontation with the air traffic controllers and Thatcher's
confrontations with the coal miners. Having an understanding
of the policy and its intents on the part of workers, unions,
and the general public can only improve the effectiveness of
the policy.
Frazer 's analysis as just outlined is very much what we
take up. it may be aligned with Friedman's approach as well.
4.2.2 Friedman's Analysis
In order to clarify whether the causation is running from
AW to AP or AP to AW, we temporarily neglect Friedman's newer
version of the Phillips curve in Monetary TrPnri. (1982). In
the old version, if the Phillips curve reflected labor supply
behavior then Friedman was insisting that the Phillips curve
was a wage bargaining relationship, m it, the workers could
at best only take into account the expected rate of inflation
in the wage bargain. Thus we have
W°  a + a ,U + a 2 P°* (4>4)
where U is unemployment rate and P 0e is expected inflation
83
rate. In words, wages are determined by unemployment (U) and
inflationary expectations (P 0e ) . Further, in his view of the
formation of inflationary expectations Friedman adopted an
adaptive framework (Friedman 1969, 124) . In it P 0e is a weight
average of previous P°, namely:
p° e = n ir M (in)'po t . If o<n<i (4 . 5)
where (ln) is the weight attached to actual inflation which
decays as one goes back from the current period into the
distant past. When n=i, P 0e =p f i>e## inflation is fully
anticipated. When n=0, expected inflation bears no relation
to the history of actual inflation rates.
Combining (4.5) and (4.4) we have
W° t = a n+a 1 U t +a 1 (ln)U t . 1 +a 2 nP° t +(ln)w° t . 1 (4.6)
Here the coefficient of P° is now a combination of the speed
of adaptive expectation n as well as the extent to which
inflationary expectations are incorporated into the wage
bargain eguation [i.e., eguation (4.4)].
Retaining his quantity theoretic relation and causation
running from money (M) to Income (Y) , Friedman reverses the
Keynesian/postKeynesian view of causation. In addition, he
explains the Phillips curve as a wage bargaining relation
where he introduce a distinction between actual and expected
rate of inflation in equation (4.4). where the Keynesians
have only the inflation rate, Friedman substitutes the
expected inflation rate. In effect we do not know the current
84
inflation rate in the current period. We have only the
expected rate. So the worker can at best take into account
the expected rate of inflation and this may be influenced by
monetary policy (Friedman, 1969).
4 2.3 The Alternative— a Restat.^pnt
As we move further into the matters of wages,
productivity, and monetary discipline as opposed to monetary
accommodation, the post Keynesians reinter the picture. They
do so by taking up a statement due to Keynes in the General
Theory (1936, 8), notably: » [It] may be the case that within
a certain range the demand for labor is for a minimum money
wage and not for a minimum real wage". Along this line the
Rousseas Weintraub's wage theorem says P = k(W/Q) (Weintraub,
1978 2830) where we have the money wage in relation to output
[i.e., W/Q = (W/L)/(Q/L) = w/q]. In addition, from the ratio
w/q, we may take a logarithm, we then have In w  Ln q and
treating each term as time rate of change (in percent) we have
d/w)(dw/dt)*l00  (l/q)(dq/dt)*100. This turns out to
coincide with the information illustrated in figures 42a and
42b. As we pointed out in the earlier sections, in order to
put monetary policy into a "sustaining" role the post
Keynesians claims that prices are a function of nominal wages,
and wages are exogenously determined by the process of
collective bargaining. They ensure that causation runs only
from A(W) to A(P), and not the reverse, but we have already
proven this approach to be flawed (appendix D) . m this
85
flawed approach monetary policy is in a "sustaining" role, and
the stock of money is endogenous. In summary, Kaldor said the
money supply cannot be exogenously determined (Kaldor 1982,
4647) .
However, in juxtaposition to all of this Keynesian/post
Keynesian approach, we have a Frazer/Friedman wage bargaining
theory, which contains elements of Frazer's analysis (sect.
4.2.1) and Friedman's (sect. 4.2.2). As introduced in
Frazer's overshooting model ( 1991a, section 12.3), with its
parallel to Friedman's treatment of transitory and permanent
components in the data series, the wage costs in labor
contracts are tied to a cost of living index and monetary
induced price changes. Via this route, wage bargaining is
determined by agents' expected rate of inflation, as in
Friedman's discussion, so we have the Frazer/Friedman wage
bargaining theory.
4  3 Testing the Hyp othesis
In this section we are going to test hypothesis 3. in
opposition to the post Keynesian's monetary accommodation
position, Frazer/Friedman provide monetary discipline and wage
adjustments to assure noninf lationary output growth at full
employment. In opposition to hypothesis 3 (sect. 1.1.2), we
have an alternative— i.e., monetary policy not only plays a
role of influencing the price level but also in influencing
the nominal wage rate and imposing some discipline on wage
bargaining process. This appears in term of labor contracts
86
which include cost of living indexes ( Frazer 1991a, sect.
12.3), and in other was which impact on the relation between
wages and productivity.
In the present testing, we once again first consider
trends (figures 1la, lib, 42a and 42b) and the related
results. We then look at the average for the split sample
(figures 44a and 44b) and the related results. In doing so
we are presently denoting the difference between the growth in
the wage and productivity in percentage as pointed out above.
For now W~  [(1/w) (dw/dt) *100]  [ (1/q) (dq/dt) *100] . This
usage with respect to the symbols parallels that shown earlier
for the monetary indicator (sect. 1. 1. l) .
4.3.1 The Trend .
Recall that in section 3.2, and figures 1la and 1lb, we
have the trend for the indicator of monetary policy for both
the U.S. and the U.K. In equations form, these trends are
M~= 2.440101 + 0.2772t ( during 1970s )
(8.456)* (2.9785)*
M~= 2.558236  0.1794t ( during 1980s )
(9.225)* (2.6246)*
for the United States; and
M~= 1.599123 + 0.1776t ( during 1970s )
(4.1175)* (2.6016)*
M~= 22.53880  1.8899t ( during 1980s )
(8.9241)* (4.5112)*
for the United Kingdom. In these equations, t is time, and
87
M~ is the indicator of monetary policy. The asterisk means
that the coefficients are significantly different from zero at
5% level of significance.
In these equations, we see a significant difference
between the monetary regimes of the 1970s and 1980s
respectively. Moreover, an empirical questions arise,
notably: If the inflation is caused by monetary forces
(accelerated money growth) , and not by nonmonetary forces (
market structure power theories of inflation ) as Rousseas
says, then we should see the time series for wages and
productivity reflecting the change in the monetary regimes.
With the governments adopting a Keynesians full employment
goal, without regard for inflation, it proceeds to accommodate
its policy to suit inflation and wages.
in contrast, in the 1980s by gradual deceleration of
growth in the money supply and achieving monetary discipline,
the governments may expect the private sector's wages to
adjust and ultimately to achieve employment up to the natural
rate. Under this regimes, the growth in wages is expected to
slow down and, if pursued long enough to become negative (
Frazer 1991a, sect. 12.3).
Now, we considers the trend line for the difference
between the growth in the wage (expressed as percentage) and
the rate of change in productivity (expressed as percentage) ,
which we shown in figure 42a and 42b for both the U.S. and
the U.K. The equations we obtained are
88
W 4.029923 + 0.4606t ( during 1970s )
(46.221)* (4.674)*
W~= 15.330757  0.7703t ( during 1980s )
(65.142)* (6.873)*
for the United States; and
W~= 2.368125 + 0.1325t ( during 1970s )
(4.462)* (3.119)*
W~ 6.103434  0.3183t ( during 1980s )
(5.387)* (7.093)*
for the United Kingdom. m these equations, w~ is the
difference between the growth in wages and the growth in
productivity [ (1/w) (dw/dt) *100  (1/q) (dq/dt) *100] f q is
output per hour of all persons, w is hourly compensation.
Considering the results just shown, we see a significant
difference between two decades for the trend lines for both
the U.S. and the U.K.. As we recall figures lia, iib, and
section 3.2, the 1970s were characterized by an upward trend
in the indicator of monetary policy for both the U.S. and the
U.K.. m the same period, the trend for the difference
between the growth in wages and the rate of change in
productivity (W~) was upward by almost 0.5 percent per annum
in the case of U.S., and the trend was upward by almost 0.15
percent per annum in U.K. m addition, the indicator of
monetary policy turns from a positive slope to a negative
slope for the 1980s in both countries (figures 1la
and 1lb; sect. 3.2) . The trend for W~ was downward by almost
89
o
o
%
20"
1968
1972
< — i — i
1976 ig Q 1984 1988^^92
Year
Figure 42a Trend Line for the Difference Between the Growth
in Wages and the Growth in Productivity U.S.
90
o
o
P
•a
&
"N.
H
O
O
S 5~
1968 1972 1976 1980 1984 1988 1992
YEAR
Figure 4 2b Trend Line for the Difference Between the Growth
in Wages and the Growth in Productivity U.K.
91
0.8 percent per annum in U.S. and by almost 0.32 percent
per annum in U.K. This shows that the change of monetary
policy, which we observe in term of the indicator of monetary
policy, has strong impact on the series for W~ which we shown
in the figure 43a and 43b. There we see the difference
between the growth in the wage and the rate of change in
productivity (w~) getting narrow when monetary policy moves
from an era of accommodative to an era of discipline. These
differences appear not only in the U.S. but also in the U.K..
Interestingly, the W~s for the U.S. and the U.K., both were
upward when the trends for the monetary indicator were upward,
and both were downward when the trends for the monetary
indicator moved into a negative direction. This is consistent
with our discussion that when the government adopts a full
employment goal without regard for inflation, there were a
wage boost when followed an accommodative monetary policies.
In contrast, under more disciplined monetary policies wage
rates adjusted downward.
The trend for W~ and M~ (figures 1la, 1lb, 42a, and 4
2b) which we found in the last two decades for both the U.S.
and the U.K. is exactly what we expect from the
Frazer/Friedman wage bargain theory, and it is unlike what we
expect from the tenet of post Keynesian economics, where the
wages are exogenously determined and in turn determine the
price level. Now, in order to verify the relationship between
92
CHANGE AT ANNUAL RATE; SEASONALLY ADJUSTED, ANNUALLY
COMPENSATION PER HOUR
PERCENT
16
1970
1975 1980
J
1985
1990
Figure 43a The Difference Between the Growth in Wages and
the Growth in Productivity, U.S.
Source. : Ferderal Reserve Bank of St. Louis.
The rate of change of hourly compensation
The growth rate of productivity
93
15%
10%
5%f
0%
5%
10%
1968
1988
1992
Year
Figure 43b The Difference Between the Growth in Wages and
the Growth in Productivity, U.K.
94
the difference between the growth in the wage [ (1/w) (dw/dt)
*100 for percent] and the growth in productivity [(1/q) (dq/dt)
*100 for percent] and the indicator of monetary policy (sect.
1.1.1), we move one step further and examine the relationship
for the 1970s and 1980s decades respectively.
Following Friedman's uses of statistical methods (sects.
2.2.1 and 2.2.2) a summary of statistical results appear in
tables 41 and 42. In addition to the simple relation
W~=a+bM~ (4.7)
, we add to the summary of results an assessment of the
interaction between the variables in the simple relation and
their interaction with time. In this respect we write two
additional equations,
W~=a+bM~+ct (4.8)
W~=a+bM~+ct+dtM~ (4.9)
and show the best fit results presented in appendix C. The
best fit results were obtained there using the "F test" which
was also discussed in the appendix C.
Applying the "F test" we have the trend lines yielding
the best fits for both countries. These "best fit" lines are
for equation (4.8) in both countries, but the coefficient c in
equation (4.8) appears along with the dummy variable t (for
the different periods; t=l for the 1970s and t=0 for the
1980s). The results obtained via this route are as follows:
W~= 0.060021 + 1.146593M"  0.020476t
95
Table 41 SUMMARY FOR FITTING MODEL BETWEEN M"
UNITED STATES
AND W"
Variables
Eauation
or Statistic
w~
=a+bM~
W~=a+bM~+ct
W"=a+bM"+ct+dtM"
Intercept
M~
t
H"t
0.055536*
0.869596*
0.060021*
1.146593*
0.020476*
0.059964*
1.151359*
0.019907*
0.021829*
R 2
SSE
TSS
0.3197
0.00411
0.00605
0.7223
0.00167
0.00605
0.7223
0.00167
0.00605
The asterisk
from zero at
means the coefficient signifi
5% level
cantly different
Table 42 SUMMARY FOR FITTING MODEL BETWEEN M~ AND W"
UNITED KINGDOM '
Variables
Eauation
or Statistic
W~=a+bM~
W"=a+bM"+ct
W~=a+bM~+ct+dtM~
Intercept
M~
t
M~t
0.024443*
0.226707*
0.025145*
0.231713*
0.001347
0.024389*
0.207438*
0.003272
0.161194*
R 2
SSE
TSS
0.6204
0.00063
0.00166
0.8449
0.00025
0.00166
8892
.00018
.00166
The asterisk
from zero at
means the coefficient signifi
5% level
cantly
different
96
in other words, we have
W~= 0.039554 + 1.146593M" ( during 1970s )
W 0.060021 + 1.146593M" ( during 1980s )
for the United States. And
W~= 0.025145 + 0.231713M~  0.001347t
in other words, we have
W~= 0.023798 + 0.231713M" ( during 1970s )
W~= 0.025145 + 0.231713M" ( during 1980s )
for the United Kingdom. In both countries we see a positive
relationship between the indicator of monetary policy (M~) and
the difference between the growth in the wage and the growth
in productivity (W~) . These results are unfavorable for the
postKeynesian view and favorable to the monetary view.
Moreover, we have a repetitive results both for U.S. and U.K.
in two different decades, and these reflect the stable
relationship in term of Friedman's definition.
Continuing to pursue the methods we introduced in section
2.3, we now turn to examining the upper and lower bounds on
the "true regression coefficient." For now, we have two
regressions. One is the difference between growth in wage and
the growth in productivity (w~) in relation to the indicator
of monetary policy (M~) , namely: W~ = a + bM~ . The other is
the indicator of monetary policy (M~) in relation to the
difference between growth in wage and the growth in
97
Table 43 THE UPPER AND LOWER BOUND ESTIMATION FOR THE
DIFFERENCE BETWEEN GROWTH IN THE WAGE AND THE GROWTH
PRODUCTIVITY AND THE INDICATOR OF MONETARY
POLICY U.S.
Variables
or Statistic
Lower
Upper
Intercept
M~
t
0.060021*
1.146593*
0.020476*
0.0413024*
2.6931674*
0.0472166*
R 2
SSE
TSS
0.7223
0.00167
0.00605
0.5602
0.00112
0.00255
means the coefficient significantly different from zero at
5% level
Table 44 THE UPPER AND LOWER BOUND ESTIMATION FOR THE
DIFFERENCE BETWEEN GROWTH IN THE WAGE AND THE GROWTH
PRODUCTIVITY AND THE INDICATOR OF MONETARY
POLICY U.K.
Variables Lower Upper
or Statistic
Intercept 0.025145* 0.0267108*
M 0.231713* 0.2819886*
t 0.001347 0.0038973
R 2 0.8449 0.8553
SSE 0.00026 0.00395
TSS 0.00166 0.02731
means the coefficient significantly different from zero at
5% level
98
productivity (W) , namely: M~ = c + d W~ . But on this
equation we perform algebraic operations (F/S 1982, 221238)
to obtain an equation W~ = a., +b 1 M" , from the results for M~
= c + d W". The two sets of results then become the upper and
lower bounds on the "true regression coefficient."
Table 43 and 44 give, in columns 1 and 2 , the
numerical estimates coefficient for these regressions for the
U.S. and U.K. respectively. For the U.S. there are:
W~= 0.039554 + 1.146593M" ( lower bound 1970s )
W~= 0.005914 + 2.693167M" ( upper bound 1970s )
W~= 0.060021 + 1.146593M" ( lower bound 1980s )
W~= 0.041302 + 2.693167M" ( upper bound 1980s )
for the U.K. the equations are:
W~= 0.023798 + 0.231713M" ( lower bound 1970s )
W~= 0.022813 + 0.281988M" ( upper bound 1970s )
W~= 0.025145 + 0.231713M  ( lower bound 1980s )
W~= 0.026710 + 0.281988M" ( upper bound 1980s )
Here the relation between the W" and the M~ either in the
lower bound estimation or in the upper bound estimation have
a positive relation. This states that monetary policy has a
positive influence on the wage rate. And the significance of
the coefficients indicates that there is a positive
correlation among the indicator of monetary policy (M~) and
the difference between the growth in the wage and the growth
in productivity (W~) in the last two decades. To be sure,
this is not supportive of the post Keynesian's position that
99
money plays no role in determining inflation and wage rates.
As long as the indicator of monetary policy was upward, which
means an accommodative monetary policy, we also find the
difference between the growth in the wage and the growth in
productivity gets wider. When the indicator of monetary
policy was downward, which means monetary discipline, the
difference between the growth in the wage and the growth in
productivity are narrowing. These results are supportive of
the Frazer/Friedman wage bargain theory and monetary view.
4.3.2 The Mean .
Parallel to the treatment in chapter 3, we again consider
figure 34a and 34b, and examine the mean for the indicator
of monetary policy for both the U.S. and the U.K., and we
recall the data summarized in table 31. m summary, the
monetary policy changed from accommodation to a period of
discipline as we move from the 1970s to the 1980s.
We see that a significant difference exists between two
policy regimes in term of the mean of the indicator of
monetary policy both in the U.S. and in the U.K.. Reflecting
upon these differences, a question arises, namely: if the
Frazer/Friedman wage bargaining theory holds and if the
inflation is caused by monetary forces rather than by non
monetary forces, then we should be able to observe changes in
the series for wage and productivity as a possible response to
the changing monetary policy. since the indicator for
monetary policy moves from a positive sign (accommodative) to
100
a negative sign (discipline) both in the U.S and the U.K., we
should observe a decline in the ratio of wages to productivity
(i.e., w/q and taking the logarithm of the ratio, as in
section 4.2.3 we have a decline in the difference between the
growth in the wage and the growth in productivity) .
Along this route, we calculate the maximum, minimum, mean
and standard deviation for the difference between the growth
in the wage and the growth in productivity. Taking note of
two different policy regimes, we find that the difference
between the growth in wages and the growth in productivity
decline not only in the U.S. but also in the U.K., as we move
from the 1970s to the 1980s. During the periods of monetary
accommodation the mean of the difference between the growth in
the wage and the growth in productivity is 6.735 percent in
the U.S. and 3.014 percent in the U.K., whereas during the
periods of monetary discipline the mean of the difference
between the growth in the wage and the growth in productivity
declines to 4.477 percent in the U.S. and 1.407 percent in the
U.K. .
These differences for the respective regimes periods are
significantly different from one another at the 1 percent
level of significance for each country.
These results are shown in figure 44a and 44b for the
U.S and the U.K., respectively. And again they are consistent
with the hypothesis advanced by Frazer and Friedman that the
inflation numbers are generated by monetary forces (namely
101
accelerated money growth) . In this respect we have referred
to Frazer/ Friedman wage bargaining theory which holds that
monetary policy has impact on the difference between the wage
rate and productivity. This view is distinct from the
hypothsis that the difference cited is caused by nonmonetary
forces (market structure power theoriesof inflation) , such as
we find in the post Keynesian economics.
4 .4 Summary
In summary, hypothesis 3 says: wage rates are determined
by productivity and market structures irrespective of monetary
policy. This we associate with a theorem due to Sidney
Weintraub and Stephen Rousseas and relate as well to the post
Keynesian view about an endogenous money supply.
Table 45 MAXIMUM, MINIMUM, MEAN AND STANDARD DEVIATION FOR
THE DIFFERENCE BETWEEN THE GROWTH IN WAGES AND THE
GROWTH IN PRODUCTIVITY
Periods of time
Min.
Max.
Mean
S.D.
United States
1970s (iregime)
1.900%
12.921%
6.735%*
3
.052%
1980s (Mregime)
0.257%
11.337%
4.477%*
3
.259%
United Kingdom
1970s (iregime)
0.104%
8.329%
3.014%*
1
767%
1980s (Mregime)
0.484%
6.081%
1.407%*
1
712%
The asterisk means of the difference between the growth in
wage and the rate of change in productivity are significantly
different from one another at the 1 percent level in each
country. 2
The tests for the significance of the difference between the
mean of the indicator of monetary policy is discussed in
appendix E.
102
5%
10%
Mean for 70s
Mean for 80s
yUri
H 1 1 1 1 h
— H 1 1 1 H
1968 1972 1976 1980
Year
I 1 H
H — i — i — i —  — h
1984 1988 1992
Figure 44a The Mean for the Difference Between the Growth in
Wages and the Growth in Productivity 1970s vs
1980s, U.S.
Mean for 70s
20%
15%
10%
5%
0%
5%
10%
103
Mean for 80s
hrfffl:^"rrfl][rh , :'nTkTdi
H 1 1 1 1 1 1 i 1 1 1 1 i 1 1 1 1 1 1 1 1 1 1
1968 1972 1976 1980 1984 1988 1992
YEAR
Figure 44b The Mean for the Difference Between the Growth in
Wages and the Growth in Productivity 1970s vs
1980s, U.K.
104
In opposition to this we have the prospect of monetary
restraint (discipline) and wage adjustments to assure
noninf lationary output growth. It is a Friedman/ Frazer view
in which the wages are determined by bargaining between the
labor and management sides of industry and the principals'
expected rate of inflation. In this Friedman/Frazer wage
bargaining theory, monetary policy plays the role of
influencing the price level and the nominal wage rate. In
this approach the prospect is left open as to whether swings
between monetary accommodation and discipline influence the
productivity of the workers ( Frazer 1991a, sect. 12.3).
We examine this issue with respect to the trends
combination of nominal wages, productivity, in the one case,
and monetary accommodation and discipline, in another case.
We also take up the differences between the 1970s and 1980s
decades by looking at average values and standard deviations
for the difference in percentage points between the wage
growth and the productivity growth, in the one case, and the
indicator of monetary accommodation (or discipline) , in the
other case.
We conclude that hypothesis 3 is rejected. Also in the
results we present in section 4.3, we observe a shared
experience in both the U.S. and the U.K. In both countries we
see a positive relationship between the indicator of monetary
policy (M~) and the difference between the growth in the wage
and the growth in productivity (W~) . Moreover, we have a
105
repetition of results both for the U.S. and the U.K. in two
different decades. This is reflecting the stable relationship
in term of Friedman's definition. As long as the indicator of
monetary policy moves upward, which means accommodative
monetary policy, we also find that the difference between the
growth in the wages and the growth in productivity moves
upward. When the indicator of monetary policy moves downward,
which means monetary discipline, the difference between the
growth in the wages and the growth in productivity also moves
downward. This shared experience in both countries again
shows the impact of monetary policy on the series we
undertake for study and this, furthermore, adds to the support
for hypothesis 1, in section 1.1.2.
CHAPTER 5
THE SAMPLE PERIODS, THE ANALYSIS OF DATA, AND THE HYPOTHESES
5. 1 Introduction
Over the post World War II years in economics there has
been for the most part the econometric method for the analysis
of time series which F/S called "the prevailing fashion in
econometric work." As reviewed by Frazer (1988, 6887) it
followed the course set by Nobel laureates Ragnar Frisch and
Jan Tinbergen and taken up by the Nobel laureate Lawrence
Klein. The major alternative which emerged to challenge this
approach has come at the hands of Milton Friedman, but it did
so as a part of Friedman's doing economics and analyzing data
for the most part, rather than from Friedman's writing about
the uses of statistical methods. 1 The arguments and debates
have been numerous and intense and widely reported with
references to the big models, reduced forms, simple models,
exogenous and endogenous variables, causation, multiple
However, we may point out that Frazer interviewed
Milton Friedman on the subject of his uses of statistical
methods after he published (with Boland) appear titled
"An Essay on the Foundations of Friedman's Methodology"
(1983). Results from the interview appear in an
unpublished document Frazer wrote with econometrician Kim
Sawyer (1984) and in Frazer (1988, chaps. 3 and 18). In
addition, Frazer studied all facts of Friedman's uses of
methods as they appear in Friedman's paper and books in
economics.
106
107
regression equation, the bounds on "true regression
coefficients," the Learner problem, the filtering of the data,
and "the prevailing fashion in econometric work" (Frazer 1973
chaps. 5 and 9; 1988, 6887 and chap 18; Frazer and Sawyer
1984) .
We make no attempt to review all of the above, although
all of the analytical problems and conflicts appear in this
dissertation. Rather than reviewing the latter, we do four
things, notably: (1) introduce four hypotheses (sect. 1.1.2),
the last of which specifically addresses the use of
statistical methods; (2) narrow the focus of controversy to
what Frazer called » the separation of effects problem" (sect.
2.3); (3) take up such crucial matters as episodes, the
filtering of time series data, and the setting of bounds on
the true regression coefficients; and (4) focus upon some time
series of a rather crucial nature as they relate to monetary
policy and overall economic performance for the U.S. and the
U.K. economies. The time series and data sources we presently
rely upon are listed in section 1.3.3.
In hypothesis 4 "the prevailing fashion in econometric
work" is said to be appropriate for the analysis of the time
series. The approach, we said, gives secondary attention to
episodes, and proceeds as if information is expected to be
obtained from an unchanging universe.
In contrast to that method, we introduce uses of
statistical methods which Frazer attributes to Friedman (1988,
108
6887). Going that route Friedman adapts a Bayesian approach
to the extent (1) that episodes move the series, (2) that
agents learn, and (3) that Friedman attempts to separate the
repetitive from the nonrepetitive or episodic part of the
time series. in addition, Friedman draws no distinction
between the agents forming expectations along classical
probability lines and otherwise having incomplete information.
The probability and the incomplete information are one and the
same and agents may view outcomes stochastically, as they
obtain new information and revise their prior view.
The role we attribute to episodes in this foregoing
context (sect. 1.2) becomes a primary distinguishing feature
in the way Friedman proceeded in the use of statistical
methods and in the way Frazer introduced Bayesian learning and
rationality on the part of economic agents (Frazer 1873,
chap. 8; 1978; and 1991a, sects. 1.1, 2.2c, 7.2c, and 14.2c).
This revision of prior view is particularly visible in the
role of psychological time (Frazer 1988, 731) which simply
gives further attention to episodes.
In confronting "the prevailing fashion in econometric
work," special analytical problems in the analysis of data are
encountered, which we summarized in chapter 1 and sections 2.2
and 2.3. But these reduce primarily to one problem, namely,
the separation of effects in the time series. In broad
outline, as taken up by Friedman and Frazer, there are special
time frames and different classes of information contained in
109
even a single time series of the sort monetary officials
confront ( chap.l and sects. 2.2 and 2.3). The time frames
are the very short run of Keynes ' s General Theory and monetary
crises as a rule, the short cycle (as delineated by the NBER's
reference dates for peaks and troughs in the transitory part
of the time series), and Friedman's long run (i.e., the trends
or permanent components in the time series) . Episodes may
enter in each of these time frames, and particularly for the
present purpose we have focused on trends— for the 1970s and
the 1980s, respectively— which we identify with distinct
approaches to monetary policy (sect. 1.2.1) and even with
different political regimes.
Going beyond the information contained in a given time
series, numerous time series may be sharing in the reactions
to episodes of the sort we point to and take up in the
discussion of exogenous and endogenous variables (sect. 1.2
and appendix A) . When this occurs, the changes in the series
are rarely independent of one another and may indeed most
commonly be responding to shared forces. Such possible
occurrences were illustrated with eguation (2.1) , section 2.3.
5 2 A Use of Conventional Method
Recall that in section 4.3.1 we examined the relationship
between the indicator of monetary policy (M~) and the
difference between the growth in wages and the growth in
productivity (W~) following Friedman's method (sect. 2.2). In
it, Friedman gives attention to the simple regression
110
technique and to the bounds on the true regression coefficient
(sects. 4.3.1 and 5.3). However, in the present section, we
take up the conventional method which we find in the work of
econometricians Hendry and Ericsson (1990). That work
proceeds as if the sampling is from an unchanging universe,
and quite separate from the time frame distinctions Friedman
draws (sect. 1.3.2). in addition, it contrast with the
analysis of the relationship between the indicator of monetary
policy and the difference between the growth in wages and the
growth in productivity.
The new results, which we obtain for the same series
analyzed earlier (section 4.3.1), are in tables 51 and 52
for the U.S. and the U.K. respectively. After the use of the
Ftest in an analysis of covariance (sect 3.2 and appendix C) ,
we found the best fit equation for each country to be
W~=a+bM~. in reference to the tables we point to the best
results for the U.S. and the U.K. respectively:
W~ = 0. 052259+0. 278580M~
W~ = 0. 037448+0. 297442M"
Comparing the results we had earlier in section 4.3.1
(table 42 and 43) with those obtained by conventional
methods and shown in tables 51 and 52, we find a low
coefficient of determination (r 2 ) for each country when using
the conventional approach and higher coefficients when using
Friedman's approach. The reasons for a low coefficient of
determination are that time series reflect the impact of
Ill
Table 51 CONVENTIONAL METHOD FOR THE RELATIONSHIP BETWEEN
THE INDICATOR OF MONETARY POLICY AND THE DIFFERENCE
BETWEEN THE GROWTH IN WAGE AND THE RATE OF CHANGE
IN PRODUCTIVITY, U.S. CHANGE
V ^ i J bl f s ~~~Eauati~an
or Statistic
W ~ =a+bM ~ W=a + bM~ + ct W~=a + bM~ + ct + dtM
Intercept 0.052259* 0.045664*'" ~ 0^045741*
M 0.278580* 0.181377* 0.196990*
l~. "" 0.015548 0.018899
~~_ — 0.116744*
R2 0.1097 0.1514 "o~.~1531~
SE 0.07512 0.07160 0.07146
*ff_ 0.08438 0.08438 0.08438
Table 52 CONVENTIONAL METHOD FOR THE RELATIONSHIP BETWEEN
THE INDICATOR OF MONETARY POLICY AND THE DIFFERENCE
BETWEEN THE GROWTH IN WAGE AND THE RATE OF CHANGF
IN PRODUCTIVITY, U.K. vnAnwi
Variables Equation"
or Statistic '
ir=a+bM ~ W=a + bM + ct W~=a + bM~ + ct + dtM~
intercept 0.037448* o'.ToTs'tT "oToTlTlT"
M 0.297442 0.046076* 0.007773
2j t "" 0.042267* 0.037677*
~__ — 0.161990
g *"" """ "™ — "" mm " m *"" m * ~ ■" ~~ *~ mm — — — — — — ~ — — _ w _
« 0.0317 0.1404 o 1591
SSE 0.94118 0.83543 0.8*1731
*ff_ 0^97196 0.97196 0.97196
frS/zero^at 5 T level* 6 Z7if ^ T ^^
112
episodes on the data series, such as we pointed to in section
1.2, and the conventional method of proceeding directly to
relate the series to one another offers no means of filtering
episodic changes out of the time series in order to focus upon
the isolation information of a more permanent sort. Thus,
while we undertake regressions for the indicator of monetary
policy (IT) and the difference between the growth in wages and
the growth in productivity (w~) using a conventional approach,
we include a lot information which is not useful for the
present purpose. Via this conventional route we cannot
achieve a clear permanent relationship that really matters for
policy making.
In considering the uses of the alternative methods, a
main question is whether Friedman's approach or H/E's yields
results which mean anything in terms of policy. Although H/E
notice a strong positive correlations between the series for
the price, the wage rate and the money stock, H/E (1990, 11
13) did not state the conclusion that wages are exogenously
determined and in turn determine the price level. Rather H/E
move one step further to examine these series in ten
approximately "equallength subsamples" of the data.
Although H/E are critical of F/S's use of "phase
averaging" ( sect. 2.2.1) in filtering the data (sect. 2.2.2)
they themselves are also engaging in data transformations in
terms of what they call "equallength subsample" as
illustrated in section 2.2.3. H/E fitted regressions to the
113
ten resulting subsamples and claimed that » ... virtually
every possible correlation between the growth rates of money
and prices can be observed" (H/E 1990, 11) . while H/E pointed
to data transformation as losing information at the earlier
time (H/E 1983, 6), now they use the method of the "equal
length subsample." The difference between this and "phase
averaging" to fit a trend is that F/S calculate phase
averaging base on the chronology data provided by the National
Bureau of Economic Research and H/E pick up their subsample
by using ambiguous, equally divided subperiod for no economic
reason. What they found (1990, figure 4a and 4b, 1113) are
results which are similar to our results in table 51 and 52
for different sample periods, but they appear to offer no
knowledge of the world that generated the data. Arbitrarily
picking up ten different subsamples, H/E obtain every possible
result, and positive and negative values for coefficient of
the indicator of monetary policy (M~) . This effort at forcing
the data into meaningless subsample provides no association
with policy making of any known sort. It is "facts without
theory," as Tjalling Koopmans once said of Friedman's early
work (Frazer 1988, 732).
5 3 A Comparison of Results
In pointing to the inappropriateness of the conventional
use of the methods, we emphasize again the superiority of
Friedman's approach. He took the view that phase averaging
separated the relevant information from information which may
114
confound estimation of the parameters. in this sense,
Friedman's use of phase averaging to obtain a trend (sect.
2.2.1) is likely to result in a more stable relation than can
be obtained from proceeding directly with the original,
unadjusted data. This would be because the positive serial
correlation within a transitory phase is reduced, and because
the effects of extreme expansions and contractions are
dampened .
Indeed, in board outline, the episodic part of the time
series, which Friedman eliminated, is picked up by H/E. They
pick up all the information F/S find irrelevant and in doing
so H/E conclude with uncertain results which provided no
positive policy associations.
As we may recall from section 4.3.1, we analyzed the
relation between the M~ and the W~ by taking up Learner's
Extreme Bounds Analysis (EBA) and F/S's use of phase averaging
in fitting trend lines and in setting the upper and lower
bounds on a "true" regression coefficient. In doing this we
obtain two regression eguations which we have pointed to
(sect. 4.3.1) .
Now, we bring forward results obtained via Friedman's
approach and juxtapose them in figures 51 and 52 with
results obtained by H/E. With Friedman's approach we obtained
upper and lower bounds for "true" regression coefficient where
for both the U.S. and the U.K. a positive relationship between
the M~ and the W~. And the true regression coefficient for
115
W"
0.100
0.080 
0.060
0.040
0.020
So 5 ufftir boMit( , tf
3oi Ljiv«T t>ownct
0.000
0.050
0.025
0.000
0.025
0.050
M"
Figure 51 Comparison for Different Results between M" and W
in the United States, Friedman vs H/E
116
0.060
0.050
0.040
0.030
0.020 +
0.010
0.000
0.200
70s upper b>0'Jnc\
■0.100
0.000
0.100
0.200
M"
Figure 52 Comparison for Different Results between M~ and W"
in the United Kingdom, Friedman vs H/E
117
the indicator of monetary policy (M~) lies between the upper
and lower bounds of 1.146593 and 2.69167 for the U.S. and
0.231713 and 0.281988 for the U.K., respectively, which we
show in figure 51 and 52 for U.S. and U.K., respectively.
Here the relation between the M~ and the w~ indicates that
monetary policy has a positive influence on the wage level
(i.e., the wage moves positively with monetary policy) . This
is very much at odds with the tenet of post Keynesian
economics which says that the wage is exogenously determined,
that it in turn determines the price level, and that monetary
policy has no influence on the price level (Rousseas 1986 74
79) .
In contrast to Friedman's approach, the results obtained
with the conventional approach have the following: lower
coefficients of determination for both the U.S. and the U.K.,
and a lower and nonsignificant regression coefficient for the
U.S.. These results, via the use of "the prevailing fashion
in econometric work" are obtained with unfiltered time series
and, in addition, take for granted an unchanging universe for
the sample period (sect. 5. 2).
5.4 Summary
In the first chapter we point to four alternative
analytical system with some claim to being positive economics.
Among them there is a wide range of differences as to
philosophy and uses of statistical methods. in this
dissertation we juxtapose them , as sufficiently distinguished
118
by outside forces to provide the prospect for significant
differences in the time series drawn from the respective
decades, to compare Friedman's economics. Moreover, as to the
four, we settle on only two of the alternatives and say that
these have the most claim to some sort of relevance in the
debates and controversies surrounding the implementation of
policies of the kind that emerge in connection with J.M.
Keynes's General Theory .
The two major alternatives are the Keynesian/post
Keynesian one and Friedman's. Furthermore, the former tends
to be most readily identified with the econometric method
passed along via Lawrence Klein and called "the prevailing
fashion in econometric work" by Friedman, and Friedman offers
his own indirect approach as embellished mainly by Frazer.
In any case, retracking these route we introduce four
hypotheses which we associate with either the Keynesian/post
Keynesian approach or Friedman's approach which extends to
rather different uses of statistical methods. They are
Hypothesis 1: The United Kingdom and the United
States have in common the same determinants of the
money demand functions (F/S 1982, sect. 5.4).
Hypothesis 2: Prices, nominal wage rates adjust
more readily in the presence of monetary
discipline.
Hypothesis — 3j_ Wage rates are determined by
productivity and market structures irrespective of
monetary policy.
Hypothesis 4: Standard econometric methods are
appropriate for analysis of the time series we deal
119
with, the hypotheses we confront, and the treatment
of episodes of the kind we encounter for the
decades of the 1970s and 1980s.
5.4. 1 Hypothesis 1
In hypothesis 1 we offer a view which F/S advanced in
Monetary Trend (1982). As we view it, their work supported
the hypothesis, but we find as well that the trends and
phenomena we considered offers further support for hypothesis
1. To be sure, we found the following: the highly similar
policy orientations of the two countries in the 1970s and
1980s decades respectively (sect. 1.2.2, 3.2 and figures 1ia,
1lb) ; similar impacts on the data series for prices and wages
(sect. 3.4.1); and shared behavior with respect to the income
velocity of money (sect. 3.2).
In contrast to the F/S view, the Keynesians and the post
Keynesians view velocity and much else in terms of the time
series as a random walk (sect. 2.2.4). However, velocity
cannot be a random walk when the two countries are sharing the
same experience with respect to it and are sharing the trends
in the monetary policy we point to.
5.4.2 Hypothesis 2
In hypotheses 2, we recall that Keynes pointed to "sticky
wage" in the 192 0s and based the General Theory on the notion
of a wage standard (i.e., that wages would remain tied to
productivity growth as total spending was managed to achieve
Keynesian full employment.) After carefully research for the
last two decades in both the United States and the United
120
Kingdom, we observed that nominal wage rates adjust more
readily in the presence of monetary discipline, as Frazer
predicted in Power and Idea (Frazer 1988, 420, 530536, 628
629) and in Alternative Analytical System ( Frazer 1991a sect.
12.3) .
Both in the U.S. and in the U.K., we observed that when
the indicator of monetary policy moved from accommodation to
discipline era the difference between the growth in wages and
the growth in productivity narrowed also. In the U.S., during
the 1970s the difference between the growth in wages and the
growth in productivity increased by almost onehalf of a
percentage point per annum, yet during the 1980s the
difference between the growth in wages and the growth in
productivity decreased by almost 0.8 percentage points per
annum. In the U.K., during the 1970s the difference between
the growth in wages and the growth in productivity increased
by almost 0.15 percentage points per annum, yet during the
1980s the difference between the growth in wages and the
growth in productivity decreased by almost 0.32 percentage
points per annum.
The similar results for the price level and nominal wage
rate (sect. 3.4.1), have given further support for hypothesis
2. In response to the accommodative monetary policy in the
1970s, we observe a positive slope for the trend line for the
price averages and the trend line for the nominal wages in
121
both the U.S. and the U.K.. And we also observe that under
the condition of monetary discipline in the 1980s the trend
lines for the price series and wage series move in a negative
direction for both the U.S. and the U.K. (sect . 3 . 4 . 1) . This
finding shows that the inflation and wage rates adjust readily
in the presence of monetary discipline. To be sure, the
inflation and wage rates for both the U.S and the U.K. reflect
the changes in monetary policy.
The foregoing findings support the second hypothesis
guite strongly, and all the results we obtained have a similar
pattern. Said differently, all the trends in the series were
upward in the 1970s and downward in the 1980s. The
implication is that the same causal force— the monetary
policy— affects the series, which underscored a stable
relationship for them. And this concept of stability is
arguably more realistic than the restrictive econometric
definition of parameter constancy (Frazer 1988, 754).
5.4.3 Hypothesis 3
In reference to hypothesis 3, we find that the post
Keynesians place monetary policy in a "sustaining" role. They
do so while arguing that prices are a function of nominal
wages, and wages is exogenously determined by the process of
collective bargaining. They presume wage compensation is
neutral with respect to the price level. m contrast,
however, we proved that this argument itself is flawed (sect.
4.3 and appendix D) . Indeed, wage rates are not determined by
122
productivity and market structures alone and irrespective of
monetary policy. Rather wage rates are affected not only by
productivity but also by the monetary policy and the inflation
rate.
In reference to the wage bargaining process, wages are
tied to labor contract and a cost of living index (Frazer
1991a, sect. 12.3). Adding Friedman's wage bargaining
eguation (Friedman 1969, 124), the workers take into account
the rate of inflation in the wage bargain, and the price level
thus gets into the process of wage bargain. As we know
already from F/S»s Monetary Trends r the price level is
strongly influenced by monetary policy. Also from the
statistical results we have with respect to the upper and
lower bounds analysis for the relation between the W~ and the
M~, we have seen that the indicator of monetary policy (M~)
moves the W~ series not only in the U.S. but also in the U.K.
This shared experience between two countries also supports
hypothesis 1.
5.4.4 Hypothesis 4
Hypothesis 4 states that standard econometric methods are
appropriate for analysis of the time series we deal with
(sect. 1.3), the hypotheses we confront (sect. 1.1.2), and the
treatment of episodes of the kind we encounter for the decades
of the 1970s and 1980s for both the U.S. and the U.K. (sect.
1.2) . For these countries we say episodes played a main role
123
in moving the time series about, and in shaping the behavior
of the units comprising the economies in guestion.
However, the results we report from using the standard
econometric methods yield both small coefficients of
determination (sect. 5.2 and tables 51, 52) and ambiguous
relations between the series that we examined. The reasons
for the smaller coefficients of determination and the
ambiguous results are (1) that time series contain changes
imposed by the impacts of episodes, and (2) that conventional
methods fail to filter episodic impact out of the data. Thus,
via the use of the fashionable methods, episodic impacts are
included. As a result, we cannot obtained a clear permanent
relationship that really matters for policy making. Taking up
H/E's uses as an example of an econometric approach
(sect. 5. 2), we find the following: lower coefficients of
determination for both the U.S. and the U.K., and a lower and
nonsignificant regression coefficient for the U.S..
In contrast to the use of "the prevailing fashion in
econometric work" approach, we have adopted Friedman's
approach which gives special attention to the role of
knowledge about the experiment which generates the time series
data and to separating out the episodic changes in the data to
find more permanent components, via this route we find more
convincing results (sect 4.3.1), namely, higher coefficient of
determination (sect. 4.3.1 and 5.2); a clear positive relation
124
between the M~ and the W~ for both the U.S. and the U.K.
(sect. 4.3.1) .
APPENDIX A
EXOGENOUS AND ENDOGENOUS VARIABLES
The problem of defining an "exogenous" variable in
economics started to gain visibility as the modern computer
came on the scene in the early 1960s. It came initially to
center about the definition of an "autonomous" variable in
economics, its relation to an "exogenous" variable (said to be
synonymous with "autonomous") , and the so called "big model"
(Frazer 1973, sect. 4.3, 5.1, 5.2, and chap. 14).
Particularly in relation to the latter, Frazer reviews
the methods of solving large system of equations (1973, app.
to chap. 14), and later (1991a, sect 3.2) says the matter of
solution for a large system of equations is not unlike that we
associate with Walras's system. There is still the notion of
as many nonredundant equations in the system as variables
within it. These "within" variables are the so called
"endogenous variables." However, there are additional notions
and analytical problems in the econometric model, where a
"policy" variable gets treated as an "exogenous" variables,
and where the multiple regression equation arises. Viewing
the multiple regression equation as follows,
y = a + a lXl + a 2 x 2 +...+ a n x n + e (1)
y is said to be regressed on x 1 , x 2 ,...,x n to obtain estimates
125
126
for the parameters (or coefficients) a Q , a,, a 2 , . . . , a n , and e
is a random error term (a term to whose value only
probabilities can be assigned; a term uncorrelated with any of
the other variables in the equation) . Along the route, an
"exogenous" variable in a stochastic model, such as equation
(1) , is a variable whose value in each period is statistically
independent of the values of the random disturbances in the
model in all periods. Moreover, the term "exogenous" may
refer to a variable in a system of structural equations or
imply the idea of an outside variables in relation to the
economic system under consideration such as the U.S. economy.
In the equation (1) instance above, there is additionally
strenuous notion that effects of the x's on y may be separated
via regression technique and revealed by the estimates for a ,
a : , a 2 ,..., a n . Where the estimated values for a Q , a 1 , a,, . . . ,
a n are not stable and where the values for the error term are
correlated with the variables within the model, the idea in
terms of structuralequationsmodel thinking is to add more
variables and equations to account for the instability. The
general idea, in other words, is to include all endogenous
variables in the structural equations model, in the work and
practice of Lawrence Klein the Keynesian and bigmodel
pioneer, we see Keynesian economics move from a twoequation
ISLM model to the 1,500 equations position of Klein's supply
demand model circa 1983 (Frazer 1984, 5153) . Such models may
or may not be excessive, depending on the "purpose" at hand.
127
In any case, the objective is not to denigrate the models.
Rather it is to consider whether economics is ready to
proclaim the usefulness of such models as far as economic
theory and policy are concerned.
Some main points are as follows: (1) given the way
economic time series (the series which correspond to the
system's inside and outside variables) move up and down
together, the clarity of exogeneity is never established [at
best we have degrees of exogeneity], and (2) the separate
effects of the variables in the regression equation are never
clearly established.
Texts authored by econometricians Maddala and Theil
respectively, give very little attention to the usesof
methods problems we point to. Even so, in connection with the
standard definitions of endogenous and exogenous variables
they say the following:
A common terminology used in econometrics for
dependent and independent variables is endogenous
and exogenous variables, respectively. Endogenous
variables are those determined within the economics
system, and exogenous variables are those given
from outside the system. (Maddala 1977, 5).
The intuitive background of this distinction is
that the values of certain variables (the exogenous
variables) are determined "from the outside," that
is, in a way which is independent of the other
(endogenous) variables are determined, jointly and
simultaneously, by the exogenous variables and the
disturbances in the way prescribed by the equations
of the system. The statistical formalization of
this idea is the assumption that the values of the
exogenous variables are stochastically independent
of disturbances of the system. (Theil 1971, 430
431).
128
Moreover, Maddala said: "An instrument is an exogenous
variables that is specifically manipulated so as to achieve
some targets (Maddala 1977, 6). Thus, "One has to use one's
judgement regarding the purpose of the investigation and the
data available to decide which variables to treat as exogenous
and which as endogenous" (Maddala 1977, 9) . In other words,
a this interpretation, the econometric technique looses its
authority on an issue of central importance to "the prevailing
fashion in econometric work" (Frazer 1988, 79; F/S 1982, 629).
APPENDIX B
METHOD OF PHASE AVERAGING
We follow standard NBER procedure in computing the phase
as a weighted average of all observation during the phase,
including both the initial and terminal points. The initial
and the terminal turning point observations are weighted one
half, The intervening observations are weighted unity. The
eguation is
0.5X 1 + S 2 n X, +0.5X n+1
n
where n = duration of phase, where unit of time is interval
between observations.
X.= observations entering into phase average, where X.
is observation at initial turning point and X , is
n+1
observation at terminal turning point.
Y = phase average.
Also we calculate the rates of change for each phase, and from
phase to other phase. The reason is obvious, relative changes
are the main subject of economic interest.
For example, we have table Bl for the income phase
average and the rates of change for income in each phase and
growth rate for income from phase to phase. Where phase
129
130
average is in billions of current dollars. Rate of change
from the initial point to the terminal point and growth rate
from phase to phase are at adjusted annual rates.
Table Bl Income Phase Average, Income Rate of Change from
The Initial Point to the Terminal Point and Income
Growth Rate from Phase to Phase
Phase ref . Quarter Phase Average Rate of Change Growth Rate
1969IVQ 
1970IVQ 
1973IVQ 
1975IQ 
1980IQ 
1980IIIQ
1981IIIQ
1970IVQ
1973IVQ
1975IQ
1980IQ
1980IIIQ
1981IIIQ
198 2IVQ
1008
1208
1472,
1949.
2687,
2940.
3144.
1982IVQ  1989IIQ
4178.2
5.4%
—
12.3%
9.9%
6.3%
10.2%
15.0%
10.8%
4.5%
13.7%
13.3%
12.5%
3.6%
6.1%
9.4%
8.4%
APPENDIX C
TESTING FOR THE BEST FIT TREND LINE
In this appendix we are going to explain how we pick up
the best fit (minimum least squares, most significant) trend
shown in sections 32, 34a, 43a and 52 rather than the
other trend lines shown in the following tables. There are
two hypotheses used in testing the relation between variables
which will determine the best fit trend line. In the present
case we are dealing with trend lines for the entire sample
period and the sample periods we select (namely the 1970s and
1980s; sect. 1.2a) and the two hypotheses are (1) that the
righthand side variables are interacting and (2) that they
are not interacting. The test of the hypotheses reduces to
one test, namely the Ftest (Frazer 1973, 42) which analyzes
and compares variance among different models. The Ftest will
assure the relation of the trend line that best fits the data.
In using the "Ftest" we undertake a comparison of
equations where one is called the complete model and the other
the reduced model. In the complete model we are adding
variables to the reduced, to determine whether the addition of
the variables improves the fit of the line (or plane) to the
data.
In the particular cases at hand, we have the following:
131
132
Set I M~=a+bt
M~=a+bt+cZ
M~=a+bt+cZ+dtZ
Set II W~=a+bt
W"=a+bt+cZ
W~=a+bt+cZ+dtZ
In these sets of equations we have M~ , the indicator of
monetary policy, W , the difference between the growth in wage
and the rate of change in productivity, t, as time, Z, a dummy
variable when data we have is picking up from the 1970s then
it equals to one otherwise it equals to zero. Plus we have
split the sample period 1970 to 1990 into two sub periods,
with the view to doing two things (1) improving the fit of the
trend lines, and (2) ultimately testing the hypothesis that
the universe for the 1970s is different from that of the 1980s
(sect. l.2a), making the comparisons within each set of
equations we are left with the best fit trend line. The "F
value" for the selection of the best fit trend line is, in
general terms,
( SSE R  SSE C )/( k  g )
SSE c /[ N  ( k+1 ) ]
where k and g are the numbers of independent variables in the
two different models, SSE R and SSE C are the sums of squared
errors for the reduced and complete models, and df 1 = k  g
and df 2 = N  (k+1) . In the specific cases where we have
obtained "F value" for the U.S. and the U.K. and for the two
133
sets of equations above, so we have four different tables
showing the statistical results obtained for the best trend
lines. Example of the "F values" obtained in arriving at the
results in the four different tables below are as follows:
Table Cl
(SSE  SSE )/(k  g) (0. 074160. 05186W1
F = . "
SSE C /[N  (k+1)] 0.05186/73
= 31.39
a large and significant value. That tells us the trend line
for the difference between the wage rate and the rate of
change in productivity is significantly different between the
last two decades, during the 1970s (monetary accommodation)
the trend was upward by almost 0.5 percent every quarter.
Yet, moving into the 1980s (monetary discipline) the trend was
downward by almost 0.8 percent every quarter, shown as figure
42a. And equation W~=a+bt+cZ+dtZ does give a significant
improvement in the fit than equations W~=a+bt+cZ, or W~=a+bt,
shown as figures Cl and C2 .
Table C2
(SSE  SSE )/(k  g) (0. 022880. 01962J/1
F = __
SSE C /[N  (k+1)] 0.01962/74
= 12.3
a large and significant value. That tells us the trend line
for the difference between the wage rate and the rate of
change in productivity is significantly different between the
last two decades. During the 1970s (monetary accommodation)
134
the trend line was upward by almost 0.15 percent every
quarter. Yet, turn into the 1980s (monetary discipline) the
trend line was downward by almost 0.3 2 percent every quarter.
And equation W~=a+bt+cZ+dtZ, shown as fiqure 42b, does qive
a siqnificant improvement in the fit than equation W~=a+bt+cZ
and W~=a+bt, shown as figures C3, C4.
Table C3
(SSE R  SSE c )/(k  g) (0. 097060. 09186) /l
F = =
SSE C /[N  (k+1)] 0.09186/74
= 4.18
a significant value. That tells us the trend line for the
indicator of monetary policy is significantly different
between 1970s and 1980s (sect. 1.2a). And equation
M~=a+bt+cZ+dtZ, as shown in figure 1la, is more fit than
equation M~=a+bt+cZ and M~=a+bt, shown as in figures C5 and
C6, for describing the trend for the indicator of monetary
policy.
Table C4
(SSE R  SSE c )/(k  g) (0. 253760'. 18804)/1
F = =
SSE C /[N  (k+1)] 0.18804/74
= 25.86
a significant value. That tells us the trend line for the
indicator of monetary policy is significantly different
between 1970s and 1980s (sect. 1.2a). And equation
M~=a+bt+cZ+dtZ, as shown in figure 1lb, is more fit than
equation M~=a+bt+cZ and M~=a+bt, shown as in figures C7 and
135
C8, for describing the trend for the indicator of monetary
policy.
The tables and figures follow.
Table Cl SUMMARY FOR FITTING MODEL FOR THE TREND LINE FOR THE
DIFFERENCE BETWEEN THE WAGES AND THE GROWTH IN
PRODUCTIVITY, U.S.
Variables
Equation
or Statistic
W"=a+bt
W"=a+bt+cZ
W~=a+bt+cZ+dtZ
Intercept
t
Z
tz

3.948316
0.1966
1.694475
0.0831
1.458256
15.330757
0.7703
11.360698
1.2309
R 2
SSE
TSS
0.1089
0.07519
0.08438
0.1211
0.07416
0.08438
0.3853
0.05186
0.08438
Table C2 SUMMARY FOR FITTING MODEL FOR THE TREND LINE OF THE
DIFFERENCE BETWEEN THE WAGES AND THE GROWTH IN
PRODUCTIVITY, U.K.
Variables
or Statistic
W"=a+bt
Equation
W =a+bt+cZ
W =a+bt+cZ+dtZ
Intercept
t
Z
tz
3.014804
•1.6061
3.425262
0.0844
0.773223
2.368125
0.1325
3.735309
0.4508
R 2
0.
1792
0.
1952
0.
3098
SSE
.02334
.02288
.01962
TSS
.02843
.02843
.02843
136
Table C3 SUMMARY FOR FITTING MODEL FOR THE TREND LINE FOR THE
INDICATOR OF MONETARY POLICY, U.S.
Variables Equation
or Statistic
M"=a+bt M~=a+bt+cZ
M~=a+bt+cZ+dtZ
Intercept
t
Z
tz
5.251395
0.2645
1.320804
0.0664
0.043035
2.558236
0.1794
0.118135
0.4566
R 2
SSE
TSS
0.1401
0.10610
0.12338
0.2133
0. 09706
0.12338
0.2554
0.09186
0,123 38
Table C4 SUMMARY FOR FITTING MODEL FOR THE TREND LINE FOR THE
INDICATOR OF MONETARY POLICY, U.K.
Variables Equation
or Statistic ~
M~=a+bt M~=a+bt+cZ
M~=a+bt+cZ+dtZ
Intercept
t
z
tz
6.455145
0.7161
7.220576
0.7767
0. 007891
22 .53880
1.8899
20.939677
2.0675
R 2
SSE
TSS
0.3328
0.25407
0.38077
0. 3336
0.25376
0.38077
0.5062
0.18804
0.38077
137
o
o
20
H
«
+»
15
&
T3
9
10
o
o
H
*
*
H
1968 1972 1976 1980^^4 WeB ' *
Year
992
Figure Cl The Trend line for w" , as W = a + bt + cZ
United States
138
Figure C2 The Trend line for W" , as W"= a + bt
United States
139
1992
Figure C3 The Trend line for W, as W a + bt + cZ
United Kingdom
140
S 20
«
4J
15
? 10
5
o
o
1 +
■p
3 5 +
~104— ♦
H h
H 1 h
H 1 h
1968 1972 1976 1980
YEAE
H 1 1 1 h
H h
1984 1988 1992
Figure C4 The Trend line for VT , as W"= a + bt
United States
141
20 I ■ ' i I i i i i , i i  , , , ,  , ,
1968 1972 1976 1980 1984 1988 1992
Year
Figure C5 The Trend line for M" , as M"= a + bt + cZ
United States
20
142
_ 10
s
o
o
H
S
o
"
10
20
■I — " — i — I — i — i — t
1968 ig72 1976
H 1 h
H 1 h
1980
Year
H 1 1 1 h
1984 1988 1992
Figure C6 The Trend line for M" , as M"= a + bt
United States
143
1992
Figure C7 The Trend line for M" , as M"= a + bt + cZ
United Kingdom
144
1992
Figure C8 The Trend line for M~ , as M" = a + bt
United States
APPENDIX D
WAGE COMPENSATION PRICE NEUTRAL ?
The post Keynesians in order to put monetary policy into
a "sustaining" roles claims that prices are a function of
nominal wages, and wage is exogenously determined by the
process of collective bargaining. Which they presume wage
compensation is price neutral. The view emerged that
compensation of money wage earners according to the average of
rate of productivity growth was neutral with respect to the
price level, ceteris paribus, where the sectoral weights
attached to the individual prices and rates of productivity
growth were the initial sectoral shares of value added. This
contention is easily demonstrated in a onesector model, even
in the presence of nonlabor (e.g., raw material or
agricultural) variables costs of production, because the
compensation rule maintains real per unit labor costs.
However, in a multisectoral model this contention does not
hold even in the absence of nonlabor variables costs of
production, this shows that the theory itself is flawed.
Along this line, we recall RousseasWeintraub wage
theorem, namely:
P = k (W/Q)
= k (W/L)/(Q/L) (1)
Which can be rewritten as the prices of consumption (C) and
145
146
investment (I) goods, that will be able to a multisectoral
analysis:
Pi  (1+U,) mj.L./Q. (i = c, I) (2)
where mw ; , L. , Q. and u ; are , respectively, the money wage per
worker, employment, real output, and the markup in sector i
(i  C, I) .
Given constant Uj then the rate of change of p. may be
written as:
p,° = mw ; ° + (VQ,) (i = c, I) ( 3 )
where the ° notation denotes the rate of change of the
variable.
Denote the price index by P where
P = sp, + (1s) p c (4)
and the constant weight s represents the initial share of
value of added produced in sector I; that is,
s = P, 1 Q, 1 /Y 1 (5)
where Y 1 denotes total value added the total output of
consumption and investment goods.
The symbol A donates the average economywide rate of
productivity change where
A = sCQj/L,) + (ls)(Q c /L c )° (6)
the same sectoral weights are used in the computation of the
price index and the average rate of productivity growth.
Utilizing (3), (4) we can show that
P° = sp [Pl °/P + (ls)p c p c °/P (7)
where the time superscripts have been erased for simplicity.
147
If the averageproductivity rule ( A = mw°) is applied then
wages in each sector are increased at rate a rate A so that
from (6) , (2) and (3)
P° = (sp l /P)[A°+(L l /Q l ) ] + [(ls)p c /P][A°+(L c /Q c ) ]
= [s(ls)/P][(Q c /L c ) (Q [ /L I )°](p I p c ) ( 8 )
The inflation rate is zero either if the prices in each sector
are equal or in the case of equal rates of productivity growth
in each sector so that the prices p,,p c are restored by the
adjustment rule, as in the single sector case. However, even
if p,,p c are equal in the initial period, different rates of
price change in the two sectors, caused by unequal rates of
labor productivity growth will ensure their subsequent
inequality. Period by period application of the adjustment
rule will be nonneutral with respect to the price level due
to the divergence of p, and p c . Hence, wage adjustment
according to economywide productivity growth is not neutral
which contradicts to the assumption of the post Keynesiaris
(Rousseas 1986, 74) with respect to the price level because
the weights applied to sectoral productivity growth rates do
not equal to the corresponding weights in the calculation of
the inflation.
APPENDIX E
TESTING FOR DIFFERENCE IN AVERAGE VALUES
FOR THE 1970S AND 1980S RESPECTIVELY: M~ AND W~
We see the value difference exists between two different
regimes in the mean of the indicator of monetary policy and
the mean of the difference between the wage rate and the
productivity rate both in the United States and in the United
Kingdom. One question comes up into our mind. Are these
differences significantly different toward one another ? in
the following sections we intend to answer this question. The
most common procedure for comparing two groups on a
characteristic measured on at least an interval scale is to
make inferences about their means and the difference between
them. Let Ml equals to mean of the indicator of monetary
policy for 1970s, and m 2 equals to mean of the indicator of
monetary policy for 1980s. We shall first consider the
situation in which the samples are obtained independently, and
the samples sizes are sufficiently large to obtain a normal
sampling distribution.
From the Central Limit Theorem, we know that if the
samples size n, of the first group is sufficiently large, the
sampling distribution of Q 1 is approximately normal about Ml
with variance S\ : = s\fn v where S\ is the population variance
for that group. Similarly, The sampling distribution of u 2 is
148
149
l 2
approximately normal about m 2 with variance 6 2 02 = S 2 /n , if n
is sufficiently large. u 2  u, , an unbiased point estimator
of m 2  M v has a sampling distribution that is approximately
normal about \i 2  M, with standard deviation
5 02oi = ^ ( 52 ui +<jl u 2 ) = V (SV n i + <J 2 2 /n 2 )
This leads us to the form for a confidence interval for n 2 ^
^2 ~ <M ± z a/2 <J 02 . Q1
As usual, we take the best point estimate of M 2  M,, and add
and subtract a zscore multiplied by the standard deviation of
the estimate. The theoretical formula for the standard
deviation involves the population variances s\ and 6 2 2 , which
are nearly always unknown in applications. in the large
sample case considered here, we can substitute the sample
variances a 2 , and a 2 2 as point estimates for s\ and S 2 2 in the
formula for S. 2 _ 01 without significantly affecting the results.
As a point estimated of m 2  jt», ( where m 2 is the mean of
the indicator of monetary policy for 1980s, and Ml is the mean
of the indicator of monetary policy for 197 0s) , the
difference in the mean of the indicator of monetary policy for
the 1970s to the 1980s in the United States, we would use u 2 
u, = 0.267%  3.388% = 3.655%. A 99% confidence interval
for n 2  y u 1 is
(3.655%) ± 2.58 V[ (1 . 760%) 2 /40 + (4 . 800%) 2 /38 ]
= 3.655% ± 2.58 V[0. 00683%]
 3.655% ± 2.132%
= ( 5.787% , 1.523% )
150
Since the confidence interval for m 2  p, contains only
negative values, we are essentially concluding that m 2 is
smaller than Ml at 99% confidence level. That means the mean
of the indicator of monetary policy for 1980s is smaller than
that for 1970s.
Also we test H Q : m 2 = ^ against H,: jt* 2 < /*,. The
alternative hypothesis reflects the mean of the indicator of
monetary policy for 1970s have larger mean, that means during
1970s monetary policy is more accommodative than 1980s. Now,
ct 02u1 = V[0. 00683%]
so that
z = [u 2 " a i]/ a u201 " [3.655%]/V[0.00683%] = 4.42
For this test, the Pvalue would be P = 0.00003. Thus, there
is substantial evidence that the indicator of monetary policy
for 1970s have larger mean than that for 1980s. Which also
means that in the 1970s monetary policy is more accommodative
than in the 1980s.
Like the previous section we test the monetary policy
indicator between 1970s and 198 0s for the United Kingdom also.
As a point estimated of M 4  M 3 ( where m 4 is the mean of the
indicator of monetary policy for 1980s, and m 3 is the mean of
the indicator of monetary policy for 1970s) , the difference in
the mean of the indicator of monetary policy for the 1970s and
the 1980s in the United Kingdom, we would use u 4  u 3 = 4.141%
 2.642% = 6.783%. A 99% confidence interval for m 4  M 3 is
(6.783%) ± 2.58 V[ (5 . 290%) 2 /40 + (4 . 287%) 2 /38 ]
151
= 6.783% ± 2.58 V[0. 01182%]
= 6.783% ± 2.806%
= ( 9.589% , 3.977% )
Since the confidence interval for ^  % contains only
negative values, we are essentially concluding that ^ is
smaller than m 3 at 99% confidence level. That means the mean
of the indicator of monetary policy for 1980s is smaller than
that for 1970s in the United Kingdom also.
And we test H Q : ^ = M3 against H a : ^ < Mj . The
alternative hypothesis reflects the mean of the indicator of
monetary policy for 1970s have larger mean, that means during
1970s monetary policy is more accommodative than 1980s. Now,
a Q4u3 = V[0. 01182%]
so that
z = [u 4  u 3 ]/ct. 4 .. 3 = [6.783%]/V[0.01182%] = 6.236
For this test, the Pvalue would be P = o. 000001 Thus, there
is substantial evidence that the indicator of monetary policy
for 1970s have larger mean than that for 1980s. which also
means that during the 1970s monetary policy is more
accommodative than monetary policy in the 1980s in the United
Kingdom.
Let m 5 eguals to mean of the difference between the
growth in wage and the rate of change in productivity for
1970s, and M6 eguals to mean of the difference between the
growth in wage and the rate of change in productivity 1980s.
As a point estimated of m 6  M 5 , the change of the mean of the
152
difference between the growth in wage and the rate of change
in productivity between the 1970s and the 1980s, we would use
u 6  u 5 = 4.477%  6.735% = 2.258%. a 99% confidence interval
for ii b  fj, 5 is
(2.258%) ± 2.58 V[ (3 . 052%) 2 /40 + (3 . 259%) 2 /38 ]
= 2.258% ± 2.58 V[0. 00511%]
= 2.258% ± 1.845%
= ( 4.103% , 0.413% )
Since the confidence interval for m 6  n, contains only
negative values, we are essentially concluding that a A is
smaller than m 5 at 99% confidence level. This means the mean
of the difference between the growth in wage and the rate of
change in productivity for 1980s is smaller than the mean of
the difference between the growth in wage and the rate of
change in productivity for 1970s in the United States. Yet,
into the 1980s these two growth rates obviously get closer.
Thus, when we make a point estimated of m 6  M 5 , we have all
negative estimated values.
Also we test H Q : Mfi = ^ against H a : /x 6 < M 5 . The
alternative hypothesis reflects the mean of the difference
between the growth in wage and the rate of change in
productivity for 1970s larger than the mean of the difference
between the growth in wage and the rate of change in
productivity for 198 0s. Now,
CT u6u5 ■ V[0. 00511%]
so that
153
z = [u 6  u 5 ]/ct. 6Q5 = [2.258%]/V[0.00511%] = 3.15
For this test, the Pvalue would be P = 0.00053. Thus, there
is substantial evidence that the difference between the growth
in wage and the rate of change in productivity for 1970s have
larger mean than the difference between the growth in wage and
the rate of change in productivity for 198 0s in the United
States.
Let M/ equals to mean of the difference between the
growth in wage and the rate of change in productivity for
1970s, and m 8 equals to mean of the difference between the
growth in wage and the rate of change in productivity 1980s.
As a point estimated of ^ a  M/ , the change in the mean of the
difference between the growth in wage and the rate of change
in productivity for the 1970s to the 1980s, we would use u 
8
u 7 = 1.407%  3.014% = 1.607%. a 99% confidence interval for
M 8  M 7 is
(1.607%) ± 2.58 V[(3.014%) 2 /40 + (1 . 407%) 2 /38 ]
= 1.607% ± 2.58 V[0. 00279%]
= 1.607% ± 1.363%
= ( 2.970% , 0.244% )
Since the confidence interval for m 8  M 7 contains only
negative values, we are essentially concluding that M „ is
smaller than M? at 99% confidence level. This means the mean
of the difference between the growth in wage and the rate of
change in productivity for 1980s is smaller than the mean of
the difference between the growth in wage and the rate of
154
change in productivity for 1970s in the United Kingdom.
Also we test H Q : H = M/ against Ha : Mfi < M? . The
alternative hypothesis reflects the mean of the difference
between the growth in wage and the rate of change in
productivity for 1970s larger than the mean of the difference
between the growth in wage and the rate of change in
productivity for 1980s. Now,
CT u807 " V[0. 00279%]
so that
z = [u 8  u 7 ]/a 08 .. 7 = [1.607%]/V[0. 00279%] = 3.04
For this test, the Pvalue would be P = 0=00083. Thus, there
is substantial evidence that the difference between the growth
in wage and the rate of change in productivity for 1970s have
larger mean than the difference between the growth in wage and
the rate of change in productivity for 1980s.
References
Allais Maurice. 1966. » a Restatement of the Quantity
??;? r 5f 1 i MOney '" Ame rican Economic Review 56 (December) :
11231157. '
Barro, Robert J. 1976. "Rational Expectations and the Role
of Monetary Policy." Journal of Monetary Bmnn^nc 2
(January): 132. s
= = ' 1977. "Unanticipated Money Growth and
Unemployment in the U.S.?" American Economic Pp.vjpw 67
March): 101115.
7^—t 1 1981 : Money, Expectations , and Busing
g ycles: Essays in Macroeconnmirfi New York . Academic
1984 • Macroeconomics. New York: John Wiley
& Sons, Inc
, and Grossman, Hershell. 1971. "A General
Disequilibrium Model of Income and Employment." Americ an
Economic Review (March) .
B ° lan ^ i ^ r " nC ? A ' i i 979 ' " A Criti <* ue °f Friedman's
Critics." Journal of Economic Literature 17 (June) : 503
522 . '
Bomberger, William, and Frazer, William. 1981. "Interest
Rates, Uncertainty and the Livingston Data." Journ al of
Finance (June): 661615. tiioj^or
Bordo, Michael D. , and Schwartz, Anna J., eds. 1984 A
Retrospective on the Clas sical Golds standard 1921193 1~
Chicago: University of Chicago Press National Bureau of
Economic Research.
Bronfenbrenner, Martin. 1976. "Elements of Stagflation
The °ry, " Zeitschrift fur Nati on a i ok ono m i e 36(1): i 8 .
155
156
Cagan Phillip. 1965. Determinan ts and Effec ts of Change in
the Stock of Money, 18751960 . New York: Columbia
University Press for the National Bureau of Economic
Desai M. J. 1981. "Inflation, Unemployment and Monetary
T«i 1CY 7xt *? Ex P erience " British Rev iew of Er.nnnrmV
Issues (Novermber) .
Martin's Pr ess 1 . 982 ' Testinc * Monetarism. New York: St.
Eckstein Otto. 1978. The Great Recession, with , p^^,^
cSm an n ' ^ Y ° rk: North Holland Publishing
Fisher, Irving. 1896. Appreciation and Tnt.Prp.t New York .
Macmillan.
Frazer, William. 1963. "Some Comments on Professor Ritter's
Interpretations of Keynes and Changes in Income
Velocity " Sc hweizerische Zeii^chrift fur
Volkswirtschaft an d Statistik 99(no i~ 7079.
1967 • The Dema nd for Money . Cleveland
Ohio: The World Publishing Company
 ■ .  . • 1968. in Compendium on Monetary Policv
Guidelines and Federal Reserve Structure. Washington,
d.c: U.S. Government Printing Office (December).
r^ • 19 "73. Crisis i n Economic Theory . Gaines
ville, Florida: University of Florida Press.
• 1978a. "Evolutionary Economics, Rational
Expectations, and Monetary Policy. " Journal of Economic
Issues 12 (June): 343372.
= .  . ,— • 1978b. "The Government Budget Constraint."
Public Finance Q uarterly 6 (July): 381387.
^^ . 1983. "Lord Kaldor, Friedman and Pertinent
Episodes." Wall Stre et Review of Books 4: 261284.
_^ — . 1984. "Fiscal Theory and Friedman." Wall
Street Review o f Books 12 (Winter) : 7679.
— . 1984. "The Economics of Supply and Demand
and Friedman." Economic Notes 13 (No. 3): 4773.
157
=c —. — • 1988  Power and Ideas: Milton Friedman .nH
the Big UTurn , 2 vols. Gainesville, Florida: Gulf^
Atlantic Publishing Company.
r = * 1991a. Monetary Economics with Alternati ve
Analyses . Gainesville, Florida: University of Florida
ir X fc: S S •
Thafffh «' 1991b. A European Monetar y System: a
Thatcher /Monet ar is t P erspective . Gainesville, Florida
University of Florida Press.
, and Boland, Lawrence. 1983. "An Essay on the
Foundtions of Friedman s Methodology." American Kmnn.i.
Review 73 (March) : 129144.
and Yohe, William. 1966. Introduction to
the Analytics and Institutions of Mone~and Banking .
Princeton, New Jersey: D. Van Nostrand Company Inc.
, and Sawyer, Kim, eds. 1984. Taped Discussion
on Some Uses of Statistical Methods with Milton Friedman
June 25, Las Vegas, Nevada. '
, and Hellier, David. 1990. "Money Demand, A
Stable Relation and Episodic Change." Southern E^onoiic
Journal 57 (October) : 234246. ^^
Friedman, Milton. 1949. "The Marshallian Demand Curve "
Journal of Political Economy 57 (December) : 463495.
 — * 1951. Test of Econometric Model , i n
Conference on Business Cycles. New York: "National
Bureau of Economic Research, Inc.
Pafoe „ ; J" 953  . Tne Case for Flexible Exchange
Rates." In Essays m Pn S1 >iv e Economic . Chicago
Illinois: University of Chicago Press.
I 957  A Theory of The Consump tion Functiojn
Princeton, New Jersey: Princeton University Press for the
National Bureau of Economic Research.
5 « . ,  ' i962  " The Interpolation of Time Series by
Related Series." Technical Paper No. 16, National Bureau
of Economic Research.
_^ — . 1968. "Factors Affecting the Level at
£«JJ ****•■•" ■ In D ° naM P  Jacobs and ^chard T.
Pratt, eds., Savings and Resid ential Financing: 1 968
Conference Proceeding s. Chicago: U.S. Savings and Loan
League 1968, 1127.
158
• 1969. The Optimum Quantity of Money and
Other Essays. Chicago: Aldin Publishing Company.
1974. "A Theoretical Framework for Monetary
Analysis." In Robert J. Gordon, ed., Milton Friedman's
Monetary Framework. Chicago: University of Chicago
Press.
Publishing Company
1976. Price Theory . Chicago: Aldin
1977. Inflation and Unemployment ,
Stockholm: The Nobel Foundation.
• 1983. "The Keynes Centenary: A Monetarist
Reflects." The Economist 287(June 4): 1719.
• 1984. "Monetary Policy for the 1980s." In
John H. Moore, ed. To Promote Prosperity: U.S. Domestic
Policy in the Mid1980s . Stanford University,
California: The Hoover Institution.
1984. In Heller, H. Robert Crockett, Andrew
Friedman, Milton, Niskansen, William A. , and Sinai Allen.
"Economic Outlook." Contemporary Policy Issues 3 (Fall);
1582.
and Roosa, Robert V. 1967. The Balance of
Payments : Free versus Fixed Exchange Rates . Washington,
D.C.: American Enterprise Institutes for Public Policy
Research.
and Schwartz, Anna J. 1963. A Monetary
History of the United States. 18671960 . Princeton,
N.J.: Princeton University Press for National Bureau of
Economic Research.
, and Schwartz, Anna J. 1970. Monetary
Statist ics of the United States . New York: Columbia
University Press for the National Bureau of Economic
Research.
/ and Schwartz, Anna J. 1982. Monetary Trends
in the United S tates and the United Kingdom . Chicago,
Illinois: University of Chicago Press for NBER.
and Schwartz, Anna. 1982. "The Effect of
the Term Structure of Interest Rates on the Demand for
Money in the United States." Journal of Political
Economy 90(No. 1): 201212.
159
Goodhart, Charles. 1982. "Monetary Trends in the United
States and the United Kingdom: ABritish Review." Journal
of Economics L iterature 20 (December): 15401551.
1988. The Evolution of Central Banks.
Cambridge, Massachusetts: MIT Press
Gould, John P., and Charles R. Nelson. 1974. "The Stochastic
Structure of the Velocity of Money," American Economic
Review (June): 756772.
Hansen, Alvin H. 1953. A Guide to K P . Y n Pfi . New York:
McGrawHill Book Company, Inc.
Harrod, Roy. 1971. "Discussion Paper." In J. C. Gilbert and
R. Sedgewick, eds. Monetary Theory and Monetary Policy in
the 1970's. Oxford: Oxford University Press.
Hass, Jane W. , and William Frazer. 1979. "The Political
Business Cycle: Adjustments through Persuasion Tech
niques." Economic NotP.s (Number 2): 453466.
Hendry David F. , and Ericsson, Neil R. 1984. "Assertion
Without Empirical Basis: An Econometric Appraisal of
Monetary T rend s in ... The United KinaHnm by Milton
Friedman and Anna Schwartz." International Finance
Discussion Papers (Number 270) Board of Governors of the
Federal Reserve System.
= — . July, 1989. "An
Econometric Analysis of UK Money Demand in Monetary
Trends in The United States and The United Kingdom by
Milton Friedman and Anna Schwartz." International
Finance Discussion Pap ers (Number 355) Board of Governors
of the Federal Reserve System (forthcoming American
Economic Review 1991) .
— — __July, 1990. "Modeling
The Demand for Narrow Money in The United States and The
United Kingdom." International Finance Discussion Papers
(Number 383) Board of Governors of the Federal Reserve
System.
Henri Theil 1971. P rinciples of Econometrics , New York: John
Wiley & Sons, Inc.
Hicks, J. R. 1939. Value an d Capital . Oxford: Oxford
University Press. Second edition 1946.
Hicks, John. 1950. A Contribution to the Theory of the
Trade Cycle. London: Oxford University Press.
160
77 v . 1 ^ 74 : The Crisis in Keynesian Kmnnminc
New York: Basic Books, Inc., Publishers.
1976. "Some Questions of Time in Economics." In
Anthony M. Tang eds. Evolution. Welfare and Time i n
Economics: Essays in Honor of Georaesoni^^ T
Lexington, Massachusetts: Lexington Books, D.C. Heath
and Company.
— . • 1983. "The Keynes Centenary." The
Economist 287(June 18): 1719. —
., and Allen, R.G.D. 1934. "A Reconsideration of
•17 r\F U^Iiia II m _•  m
the Theory of Value." Economica i(no. 1): 5276
?™™i ? V / n D * 1984 ' " TW ° Types of Monetarism:
Journal of Economic Literstnrp 22 (March) : 5876.
Jazwinski, Andrew H. 1970. stochastic Process^ ,nH
Filtering Theory . New York: Academic Press. "
Johnston, J. 1984. Econometri c Methods , 3rd ed. New York
McGrawHill Book Company.
Judd, John P., and Scadding, John L. 1982. "The Search for
a stable Money Demand Function: A Survey of the Post
1973 Literature." Journal of Economic Literatur e 20
(September): 15771590.
Kaldor, Nicholas. 1970. "The New Monetarism" Lloyds Ba nk
Review (July) : 1218. Y nK
"Memorandum by Professor Lord Kaldor."
— «. ~ 2 ±iuj.cDoui luiu j\aiaor.
Memoranda on Monetary Polic y , House of Commons, Treasury
and Civil Service Committee, Session 19791980. larfcrr
Her Majesty's Stationary Office.
nvf n ^ . — 7Tz — 3W; 19 ? 2 " T he Scourc r e of MgnefcarigBi
Oxford: Oxford University Press.
Keynes, John Maynard. 1923. A Tract on M onetary R.fnr,
London: Macmillan and Co., Limited.
. 1925. "The Economic Conseguence of Mr.
Churchill." Reprinted in J.M. Keynes (1932, 244270)
1930. A Treatise on Money:
The
Applied Theory of Money , Vol. II. New York: Harcourt
Brace and Company. '
n~rr, ■ 1932  " The Conseguence to the Banks of
the Collapse of Money Values (Aug, 1931)" in Essays i n
Persuasion . New York: Harcourt, Brace and Com^ny:
161
— ___, • 1932. "The Return to the Gold
Standard," m Essays In Per suasion . New York: Harcourt
Bruce and Company. '
— r • 1936. The General Theory of
Employment, Interest and Mnnpy . New York: Harcourt
Brace and Company. '
Klein, Lawrence R. 1950. The Kevnesian Revolnt.inn . The
Macmillan Company, New York.
1962. An Introduction to Econometrics.
Englewood Cliffs, N.J.: PrenticeHall, Inc.
1983. The Econ omics of SupdIv and Demand .
Baltimore, MD. : Johns Hopkins University Press
Lavoie, Marc. "The Endogenous Flow of Credit and the Post
Keynesian Theory of Money. Journal of Economic issues
18 (September) : 771797. ' ~~
Learner, Edward E. 1978. Specification Search^ . New Y ork
John Wiley & Sons.
1985. "Sensitivity Analysis Would Help."
American Econom ic Review 75 (June): 308313.
LeighPemberton, Robin. 1986. "Financial Change and Broad
Money," Lecture delivered at Loughborough University
Banking Center, 2 2 October.
^ . 1987. "The Instruments of Monetary
Policy," Mais Lecture given at City University Business
School, 13 May.
Leijonhufvud, Axel. 1968. On Kevnesian Economics and the
Economics of Keynes . New York: Oxford University Press.
Lucas, Robert E. , Jr. and Rapping, Leonard A. 1969. "Real
Wages, Employment and Inflation." Journal of Po litica]
Economy 77 (September): 721754.
— i and Sargent, Thomas J., eds. 1981.
Rational Expectations an d Econometric Practice , vols 1
and 2. Minneapolis, Minnesota: University of Minnesota
Press.
Maddala G.S., 1977. Econometrics. New York: McGrawHill
Company .
162
McAleer, M. , Pagan, A.R. , and Volker, P. A. 1978. "What Will
Take The Con Out of Econometrics?" American Economic
Review 75(June): 293307.
More, Geoffrey H. 1983. Business Cycles. Inflation and
Forecasting, second edition. Cambridge Mass. : Ballinger
Publishing Company for the National Bureau of Economic
Research, Inc.
Patinkin, Don. 1956. Money. Interest, and Prices: An
Integrating Mon etary and Value Theory . Evanston, IL:
Pow, Peterson and Company.
• 1976. Keynes' Monetary Thought: A Study of
Its Development. Durham, North Carolina.: Duke
University Press.
• 1978. "The Process of Writing The General
Theory: A Critical Survey." In Don Patinkin and J.
Clark Leith, eds. 1978. Kevnes. Cambridge, and the
General Theory. Toronto: University of Toronto Press.
, and J  Clark Leith, eds. 1978. Kevnes. Cam
bridge, and the G eneral Theory — The process of criticism
and discussion connected with the development of The
General Theory. Toronto: University of Toronto Press.
Poole, William. 1988. "Monetary Policy Lessons of Recent
Inflation and Distribution. Journal of Economin
Perspectives r 2 (Summer): 786801.
Raj, Baldev, and Siklos, Pierre L. 1987. "Keynesian and
Friedman Views of the OutputInflation Relation: The
Canadian Data, 19531982." manuscript.
Report, Delors. 1989. Report on Economic and Monetary Union
A2 — the — European Community . LuxembourgBruxelles:
European Documents.
Rotheim, Roy J. 1988. "Keynes and the Languages of
Probability and Uncertainty." Journal of Post Keynesian
Economics ll(Fall): 8299.
Rousseas, Stephen. 1982. The Political Economy of
Reaganomics: A Critique . New York: M. E. Sharpe Inc.
1986. Post Keynesian Monetary Economics . New
York: M. E. Sharpe Inc
163
Samuelson, Paul A. 1939. "Interaction between the Multiplier
Analysis and the Principle of Acceleration." Review of
Economics Statistics, 21 (May) 7578. Reprinted in 1944.
Readings in Business Cycle Theory. Philadelphia:
The Blackiston Company.
1976. Economics , Tenth Edition. New
York: McGrawHill Book Company.
Samuelson, P. A. and Solow, R. M. 1960. "Analytical Aspects
of AntiInflation Policy" American Economic Review (May) :
177194.
Sargent, Thomas J. 1987. Macroeconomic Theory , 2nd edition.
Orlando, Florida: Harcourt Brace Jovanovich, Publishers.
, and Wallace, Neil. 1976a. "Rational
Expectations and the Theory of Economic Policy." Studies
in — Monetar y Economics 2 . Federal Reserve Bank of
Minneapolis: 1633.
/ and . 1976b. "Rational
Expectations and the Theory of Economic Policy, Part II:
Arguments and Evidence," Studies in Monetary Economics 2 '.
Federal Reserve Bank of Minneapolis: 2246.
Stein, Herbert. 1984. Presidential Economics: The Making of
Economic Policy from Roosevelt to Reacran and Beyond . New
York: Simon and Schuster.
Terrell, William T., and William Frazer. 1972. "Interest
Rates, Portfolio Behavior, and Marketable Government
Securities." Journal of Finance (March): 273289.
Theil, Henri. 1971. Principles of Econometrics . New York:
John Wiley & Sons, Inc.
Thygesen, Neils. 1989. "The Delors Report and European
Economic and Monetary Union." International Affairs 65
(August): 637652.
Tobin, James. 1982. "The Reagan Economic Plan: SupplySide,
Budget, and Inflation." In Richard H. Fink, ed. Supply
Side Economic s: A Critical Appraisal . Frederick,
Maryland: University Publications of America, Inc.
Townsend, Robert M. 1978. "Market Anticipations, Rational
Expectations, and Bayesian Analysis." International
Economic Review (June): 351378.
164
Ulmer, Melville, J. 1979. "Old and New Fashions in
Employment and Inflation Theory." Journal of Economic
Issues 13 (March): 129152.
United States Senate. 1975. First Report on the Conduct on
Monetary Policy, from the Committee on Banking, Housing
and Urban Affairs, United States Senate A Report filed
Pursuant to House Concurrent Resolution 13 3 . Washington,
D.C.: U.S. Government Printing Office.
Vickrey, William. 1954. "Stability Through Inflation." In
Kenneth K. Kurihara, ed. Post Keynesian Economics . New
Brunswick, New Jersey: Rutgers University Press.
Working, Elmer J. 1927. "What Do Statistical 'Demand Curces'
Show?" Quarterly Journal of Economics , 41 (February
1927): 221235.
Weintraub, Sidney. 1973. Keynes and the Monetarists and
Other Essays . New Brunswick, New Jersey: Rutgers
University Press.
1978. Keynes. Keynesians, and Monetarists .
Philadelphia: University of Pennsylvania Press,
1978. Capitalism's Inflation and
Unemployment Crisis . Massachusetts: AddisonWesley,
Wible, James R. 1984. "The Instrumentalism of Dewey and
Friedman." Journal of Economic Issues 18 (December) :
10491070.
Zarnowitz, Victor. 1985. "Recent Work on Business Cycles in
Historical Perspective." Journal of Economic Literature
23 (no. 2) : 523580.
BIOGRAPHICAL SKETCH
Peng Cheng Wen was born in Taiwan, Republic of China.
He received his undergraduate degree in economics at the
National Taiwan University in 1980. He is married, and 32
years old. In this coming spring, he is to be an assistant
professor in economics at Pacific College in Fresno,
California.
165
., I certify that I have read this study and that in my opinion
is ?u?lv r a m d S .^^ CCe P table Standards of scholarly presentation anS
is fully adequate, in scope and quality, as a dissertation for the
degree of Doctor of Philosophy.
j^Wb^ /Ad/Cm i
William F
Professor
er, Chairm
Economics
I certify that I have read this study and that in my opinion
H t^i^S acce P table standards of scholarly presentation anS
SiilJT '/ n u SCOpe and S ual ity, as a dissertation for the
degree of Doctor of Philosophy.
Berg
Professor of Economy
i*. ^^L Certi £ y that I have read this study and that in my opinion
4« ?,??!,? acceptable standards of scholarly presentation and
h™: Y /n eq r ' in . S . C ° pe and 9 ualit y' as a dissertation for the
degree of Doctor of Philosophy.
f^crvKq
Leonard Cheng
Associate Professor of Econol
I certify that I have read this study and that in my opinion
it conforms to acceptable standards of scholarly presentation and
is fully adequate, m scope and quality, as a dissertation for the
degree of Doctor of Philosophy.
Mark Yang
Professor
This dissertation was submitted to the Graduate Faculty of the
Department of Economics in the College of Business Administration
and to the Graduate school and was accepted as partial fulfillment
of the requirements for the degree of Doctor of Philosophy
May, 1991
Dean, Graduate School
D